Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.134
1.134 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.133 2006/06/05 08:11:10 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;
1.1 noro 1595:
1.23 noro 1596: g = 0; d = 0;
1597: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1598: d = update_pairs(d,g,i);
1599: g = update_base(g,i);
1600: }
1601: sugar = 0;
1602: while ( d ) {
1603: again:
1604: l = nd_minp(d,&d);
1.14 noro 1605: if ( SG(l) != sugar ) {
1.122 noro 1606: if ( ishomo ) {
1607: stat = do_diagonalize(sugar,m);
1608: if ( !stat ) {
1609: NEXT(l) = d; d = l;
1610: d = nd_reconstruct(0,d);
1611: goto again;
1612: }
1613: }
1.14 noro 1614: sugar = SG(l);
1.71 noro 1615: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1616: }
1.53 noro 1617: stat = nd_sp(m,0,l,&h);
1.1 noro 1618: if ( !stat ) {
1619: NEXT(l) = d; d = l;
1.103 noro 1620: d = nd_reconstruct(0,d);
1.1 noro 1621: goto again;
1622: }
1.41 noro 1623: #if USE_GEOBUCKET
1.69 noro 1624: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1625: #else
1.69 noro 1626: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1627: #endif
1.1 noro 1628: if ( !stat ) {
1629: NEXT(l) = d; d = l;
1.103 noro 1630: d = nd_reconstruct(0,d);
1.1 noro 1631: goto again;
1632: } else if ( nf ) {
1.27 noro 1633: if ( checkonly ) return 0;
1.71 noro 1634: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1635: nd_removecont(m,nf);
1.118 noro 1636: if ( !m && nd_nalg ) {
1637: nd_monic(0,&nf);
1638: nd_removecont(m,nf);
1.117 noro 1639: }
1.63 noro 1640: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1641: nh = ndv_newps(m,nfv,0);
1.1 noro 1642: d = update_pairs(d,g,nh);
1643: g = update_base(g,nh);
1644: FREENDP(l);
1645: } else {
1.71 noro 1646: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1647: FREENDP(l);
1648: }
1649: }
1.77 noro 1650: if ( nd_demand )
1651: for ( t = g; t; t = NEXT(t) )
1652: BDY(t) = (pointer)ndv_load((int)BDY(t));
1653: else
1654: for ( t = g; t; t = NEXT(t) )
1655: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1656: return g;
1657: }
1658:
1.122 noro 1659: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1660: {
1661: int i,nh,stat;
1662: NODE r,g,t;
1663: ND h,nf,nfq,s,head;
1664: NDV nfv,nfqv;
1.113 noro 1665: Q q,den,num;
1666: union oNDC dn;
1.91 noro 1667:
1668: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1669: /* for nd_ps */
1670: s = ndvtond(m,nd_ps[i]);
1671: s = nd_separate_head(s,&head);
1.122 noro 1672: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1673: if ( !stat ) return 0;
1.91 noro 1674: nf = nd_add(m,head,nf);
1675: ndv_free(nd_ps[i]);
1676: nd_ps[i] = ndtondv(m,nf);
1677: nd_free(nf);
1678:
1679: /* for nd_ps_trace */
1680: if ( nd_demand )
1681: nfv = ndv_load(i);
1682: else
1683: nfv = nd_ps_trace[i];
1684: s = ndvtond(0,nfv);
1685: s = nd_separate_head(s,&head);
1.122 noro 1686: stat = nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1687: if ( !stat ) return 0;
1.113 noro 1688: NTOQ(NM(dn.z),SGN(dn.z),num);
1689: mulq(HCQ(head),num,&q); HCQ(head) = q;
1690: if ( DN(dn.z) ) {
1691: NTOQ(DN(dn.z),1,den);
1692: nd_mul_c_q(nf,den);
1693: }
1.91 noro 1694: nf = nd_add(0,head,nf);
1695: ndv_free(nfv);
1696: nd_removecont(0,nf);
1697: nfv = ndtondv(0,nf);
1698: nd_free(nf);
1699: nd_bound[i] = ndv_compute_bound(nfv);
1700: if ( nd_demand ) {
1701: ndv_save(nfv,i);
1702: ndv_free(nfv);
1703: } else
1704: nd_ps_trace[i] = nfv;
1705: }
1.122 noro 1706: return 1;
1.91 noro 1707: }
1708:
1.118 noro 1709: static struct oEGT eg_invdalg;
1710: struct oEGT eg_le;
1711:
1.82 noro 1712: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1713: {
1714: int i,nh,sugar,stat;
1.23 noro 1715: NODE r,g,t;
1.20 noro 1716: ND_pairs d;
1717: ND_pairs l;
1.82 noro 1718: ND h,nf,nfq,s,head;
1.63 noro 1719: NDV nfv,nfqv;
1.113 noro 1720: Q q,den,num;
1721: union oNDC dn;
1.118 noro 1722: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1723: int diag_count = 0;
1.20 noro 1724:
1.118 noro 1725: init_eg(&eg_monic);
1726: init_eg(&eg_invdalg);
1727: init_eg(&eg_le);
1.23 noro 1728: g = 0; d = 0;
1729: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1730: d = update_pairs(d,g,i);
1731: g = update_base(g,i);
1732: }
1733: sugar = 0;
1734: while ( d ) {
1735: again:
1736: l = nd_minp(d,&d);
1737: if ( SG(l) != sugar ) {
1.130 noro 1738: #if 1
1.122 noro 1739: if ( ishomo ) {
1740: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1741: diag_count = 0;
1.122 noro 1742: if ( !stat ) {
1743: NEXT(l) = d; d = l;
1744: d = nd_reconstruct(1,d);
1745: goto again;
1746: }
1747: }
1.130 noro 1748: #endif
1.20 noro 1749: sugar = SG(l);
1.71 noro 1750: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1751: }
1.53 noro 1752: stat = nd_sp(m,0,l,&h);
1.20 noro 1753: if ( !stat ) {
1754: NEXT(l) = d; d = l;
1.103 noro 1755: d = nd_reconstruct(1,d);
1.20 noro 1756: goto again;
1757: }
1.41 noro 1758: #if USE_GEOBUCKET
1.53 noro 1759: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1760: #else
1.69 noro 1761: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1762: #endif
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: } else if ( nf ) {
1.77 noro 1768: if ( nd_demand ) {
1769: nfqv = ndv_load(nd_psn);
1770: nfq = ndvtond(0,nfqv);
1771: } else
1772: nfq = 0;
1773: if ( !nfq ) {
1774: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1775: NEXT(l) = d; d = l;
1.103 noro 1776: d = nd_reconstruct(1,d);
1.77 noro 1777: goto again;
1778: }
1.70 noro 1779: }
1.20 noro 1780: if ( nfq ) {
1.70 noro 1781: /* m|HC(nfq) => failure */
1.113 noro 1782: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1783:
1.71 noro 1784: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1785: if ( nd_nalg ) {
1786: /* m|DN(HC(nf)^(-1)) => failure */
1787: get_eg(&egm0);
1788: if ( !nd_monic(m,&nfq) ) return 0;
1789: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1790: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1791: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1792: } else {
1793: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1794: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1795: }
1.77 noro 1796: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1797: if ( ishomo && ++diag_count == diag_period ) {
1798: diag_count = 0;
1799: stat = do_diagonalize_trace(sugar,m);
1800: if ( !stat ) {
1801: NEXT(l) = d; d = l;
1802: d = nd_reconstruct(1,d);
1803: goto again;
1804: }
1805: }
1.20 noro 1806: d = update_pairs(d,g,nh);
1807: g = update_base(g,nh);
1808: } else {
1.71 noro 1809: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1810: }
1811: } else {
1.71 noro 1812: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1813: }
1814: FREENDP(l);
1815: }
1.77 noro 1816: if ( nd_demand )
1817: for ( t = g; t; t = NEXT(t) )
1818: BDY(t) = (pointer)ndv_load((int)BDY(t));
1819: else
1820: for ( t = g; t; t = NEXT(t) )
1821: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1822: if ( nd_nalg ) {
1823: print_eg("monic",&eg_monic);
1824: print_eg("invdalg",&eg_invdalg);
1825: print_eg("le",&eg_le);
1826: }
1.20 noro 1827: return g;
1828: }
1829:
1.23 noro 1830: int ndv_compare(NDV *p1,NDV *p2)
1831: {
1.34 noro 1832: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1833: }
1834:
1835: int ndv_compare_rev(NDV *p1,NDV *p2)
1836: {
1.34 noro 1837: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1838: }
1839:
1.61 noro 1840: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1841: {
1.69 noro 1842: int i,n,stat;
1843: ND nf,g,head;
1.23 noro 1844: NODE t,a0,a;
1.113 noro 1845: union oNDC dn;
1.69 noro 1846: NDV *w;
1.113 noro 1847: Q q,num,den;
1.23 noro 1848:
1.69 noro 1849: n = length(f);
1850: #if 0
1851: w = (NDV *)ALLOCA(n*sizeof(NDV));
1852: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1853: qsort(w,n,sizeof(NDV),
1854: (int (*)(const void *,const void *))ndv_compare);
1855: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1856: #endif
1.128 noro 1857: ndv_setup(m,0,f,0);
1.69 noro 1858: for ( i = 0; i < n; ) {
1859: g = ndvtond(m,nd_ps[i]);
1860: g = nd_separate_head(g,&head);
1861: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1862: if ( !stat )
1.103 noro 1863: nd_reconstruct(0,0);
1.69 noro 1864: else {
1.71 noro 1865: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1866: if ( !m ) {
1.113 noro 1867: NTOQ(NM(dn.z),SGN(dn.z),num);
1868: mulq(HCQ(head),num,&q); HCQ(head) = q;
1869: if ( DN(dn.z) ) {
1870: NTOQ(DN(dn.z),1,den);
1871: nd_mul_c_q(nf,den);
1872: }
1.85 noro 1873: }
1.69 noro 1874: nf = nd_add(m,head,nf);
1875: ndv_free(nd_ps[i]);
1.24 noro 1876: nd_removecont(m,nf);
1.69 noro 1877: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1878: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1879: i++;
1.23 noro 1880: }
1881: }
1.71 noro 1882: if ( DP_Print ) { printf("\n"); }
1.23 noro 1883: for ( a0 = 0, i = 0; i < n; i++ ) {
1884: NEXTNODE(a0,a);
1.69 noro 1885: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1886: }
1887: NEXT(a) = 0;
1888: return a0;
1889: }
1890:
1.1 noro 1891: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1892: {
1893: ND_pairs d1,nd,cur,head,prev,remove;
1894:
1895: if ( !g ) return d;
1896: d = crit_B(d,t);
1897: d1 = nd_newpairs(g,t);
1898: d1 = crit_M(d1);
1899: d1 = crit_F(d1);
1.55 noro 1900: if ( do_weyl )
1901: head = d1;
1902: else {
1903: prev = 0; cur = head = d1;
1904: while ( cur ) {
1905: if ( crit_2( cur->i1,cur->i2 ) ) {
1906: remove = cur;
1907: if ( !prev ) head = cur = NEXT(cur);
1908: else cur = NEXT(prev) = NEXT(cur);
1909: FREENDP(remove);
1910: } else {
1911: prev = cur; cur = NEXT(cur);
1912: }
1.1 noro 1913: }
1914: }
1915: if ( !d )
1916: return head;
1917: else {
1918: nd = d;
1.34 noro 1919: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1920: NEXT(nd) = head;
1921: return d;
1922: }
1923: }
1924:
1925: ND_pairs nd_newpairs( NODE g, int t )
1926: {
1927: NODE h;
1.61 noro 1928: UINT *dl;
1.34 noro 1929: int ts,s;
1.1 noro 1930: ND_pairs r,r0;
1931:
1.20 noro 1932: dl = DL(nd_psh[t]);
1.34 noro 1933: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1934: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1935: NEXTND_pairs(r0,r);
1936: r->i1 = (int)BDY(h);
1937: r->i2 = t;
1.20 noro 1938: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1939: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1940: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1941: }
1942: NEXT(r) = 0;
1943: return r0;
1944: }
1945:
1946: ND_pairs crit_B( ND_pairs d, int s )
1947: {
1948: ND_pairs cur,head,prev,remove;
1.61 noro 1949: UINT *t,*tl,*lcm;
1.1 noro 1950: int td,tdl;
1951:
1952: if ( !d ) return 0;
1.20 noro 1953: t = DL(nd_psh[s]);
1.1 noro 1954: prev = 0;
1955: head = cur = d;
1.61 noro 1956: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1957: while ( cur ) {
1958: tl = cur->lcm;
1959: if ( ndl_reducible(tl,t)
1.20 noro 1960: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1961: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1962: remove = cur;
1963: if ( !prev ) {
1964: head = cur = NEXT(cur);
1965: } else {
1966: cur = NEXT(prev) = NEXT(cur);
1967: }
1968: FREENDP(remove);
1969: } else {
1.34 noro 1970: prev = cur; cur = NEXT(cur);
1.1 noro 1971: }
1972: }
1973: return head;
1974: }
1975:
1976: ND_pairs crit_M( ND_pairs d1 )
1977: {
1978: ND_pairs e,d2,d3,dd,p;
1.61 noro 1979: UINT *id,*jd;
1.1 noro 1980:
1981: for ( dd = 0, e = d1; e; e = d3 ) {
1982: if ( !(d2 = NEXT(e)) ) {
1983: NEXT(e) = dd;
1984: return e;
1985: }
1.34 noro 1986: id = LCM(e);
1.1 noro 1987: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1988: p = NEXT(d2);
1989: jd = LCM(d2);
1990: if ( ndl_equal(jd,id) )
1991: ;
1992: else if ( TD(jd) > TD(id) )
1.1 noro 1993: if ( ndl_reducible(jd,id) ) continue;
1994: else ;
1.34 noro 1995: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1996: NEXT(d2) = d3;
1997: d3 = d2;
1998: }
1999: NEXT(e) = dd;
2000: dd = e;
2001: continue;
2002: /**/
2003: delit: NEXT(d2) = d3;
2004: d3 = d2;
2005: for ( ; p; p = d2 ) {
2006: d2 = NEXT(p);
2007: NEXT(p) = d3;
2008: d3 = p;
2009: }
2010: FREENDP(e);
2011: }
2012: return dd;
2013: }
2014:
2015: ND_pairs crit_F( ND_pairs d1 )
2016: {
2017: ND_pairs rest, head,remove;
2018: ND_pairs last, p, r, w;
2019: int s;
2020:
2021: for ( head = last = 0, p = d1; NEXT(p); ) {
2022: r = w = equivalent_pairs(p,&rest);
1.14 noro 2023: s = SG(r);
1.1 noro 2024: w = NEXT(w);
2025: while ( w ) {
2026: if ( crit_2(w->i1,w->i2) ) {
2027: r = w;
2028: w = NEXT(w);
2029: while ( w ) {
2030: remove = w;
2031: w = NEXT(w);
2032: FREENDP(remove);
2033: }
2034: break;
1.14 noro 2035: } else if ( SG(w) < s ) {
1.1 noro 2036: FREENDP(r);
2037: r = w;
1.14 noro 2038: s = SG(r);
1.1 noro 2039: w = NEXT(w);
2040: } else {
2041: remove = w;
2042: w = NEXT(w);
2043: FREENDP(remove);
2044: }
2045: }
2046: if ( last ) NEXT(last) = r;
2047: else head = r;
2048: NEXT(last = r) = 0;
2049: p = rest;
2050: if ( !p ) return head;
2051: }
2052: if ( !last ) return p;
2053: NEXT(last) = p;
2054: return head;
2055: }
2056:
2057: int crit_2( int dp1, int dp2 )
2058: {
1.20 noro 2059: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2060: }
2061:
1.40 noro 2062: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2063: {
2064: ND_pairs w,p,r,s;
1.61 noro 2065: UINT *d;
1.1 noro 2066:
2067: w = d1;
1.34 noro 2068: d = LCM(w);
1.1 noro 2069: s = NEXT(w);
2070: NEXT(w) = 0;
2071: for ( r = 0; s; s = p ) {
2072: p = NEXT(s);
1.34 noro 2073: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2074: NEXT(s) = w; w = s;
1.1 noro 2075: } else {
1.39 noro 2076: NEXT(s) = r; r = s;
1.1 noro 2077: }
2078: }
2079: *prest = r;
2080: return w;
2081: }
2082:
2083: NODE update_base(NODE nd,int ndp)
2084: {
1.61 noro 2085: UINT *dl, *dln;
1.1 noro 2086: NODE last, p, head;
2087:
1.20 noro 2088: dl = DL(nd_psh[ndp]);
1.1 noro 2089: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2090: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2091: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2092: p = NEXT(p);
2093: if ( last ) NEXT(last) = p;
2094: } else {
2095: if ( !last ) head = p;
2096: p = NEXT(last = p);
2097: }
2098: }
2099: head = append_one(head,ndp);
2100: return head;
2101: }
2102:
2103: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2104: {
2105: ND_pairs m,ml,p,l;
1.61 noro 2106: UINT *lcm;
1.33 noro 2107: int s,td,len,tlen,c,c1;
1.1 noro 2108:
2109: if ( !(p = NEXT(m = d)) ) {
2110: *prest = p;
2111: NEXT(m) = 0;
2112: return m;
2113: }
1.14 noro 2114: s = SG(m);
1.77 noro 2115: if ( !NoSugar ) {
2116: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2117: if ( (SG(p) < s)
2118: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2119: ml = l; m = p; s = SG(m);
2120: }
2121: } else {
2122: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2123: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2124: ml = l; m = p; s = SG(m);
2125: }
2126: }
1.1 noro 2127: if ( !ml ) *prest = NEXT(m);
2128: else {
2129: NEXT(ml) = NEXT(m);
2130: *prest = d;
2131: }
2132: NEXT(m) = 0;
2133: return m;
2134: }
2135:
1.63 noro 2136: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2137: {
1.69 noro 2138: int msugar,i;
1.63 noro 2139: ND_pairs t,dm0,dm,dr0,dr;
2140:
2141: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2142: if ( SG(t) < msugar ) msugar = SG(t);
2143: dm0 = 0; dr0 = 0;
1.69 noro 2144: for ( i = 0, t = d; t; t = NEXT(t) )
2145: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2146: if ( dm0 ) NEXT(dm) = t;
2147: else dm0 = t;
2148: dm = t;
1.69 noro 2149: i++;
1.63 noro 2150: } else {
2151: if ( dr0 ) NEXT(dr) = t;
2152: else dr0 = t;
2153: dr = t;
2154: }
2155: NEXT(dm) = 0;
2156: if ( dr0 ) NEXT(dr) = 0;
2157: *prest = dr0;
2158: return dm0;
2159: }
2160:
1.77 noro 2161: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2162: {
1.3 noro 2163: int len;
1.13 noro 2164: RHist r;
1.20 noro 2165: NDV b;
1.3 noro 2166:
1.1 noro 2167: if ( nd_psn == nd_pslen ) {
2168: nd_pslen *= 2;
1.11 noro 2169: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2170: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2171: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2172: nd_bound = (UINT **)
2173: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2174: }
1.39 noro 2175: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2176: nd_ps[nd_psn] = a;
1.39 noro 2177: if ( aq ) {
1.63 noro 2178: nd_ps_trace[nd_psn] = aq;
2179: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2180: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2181: } else {
1.63 noro 2182: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2183: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2184: }
1.77 noro 2185: if ( nd_demand ) {
2186: if ( aq ) {
2187: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2188: nd_ps_trace[nd_psn] = 0;
2189: } else {
2190: ndv_save(nd_ps[nd_psn],nd_psn);
2191: nd_ps[nd_psn] = 0;
2192: }
2193: }
1.1 noro 2194: return nd_psn++;
2195: }
2196:
1.128 noro 2197: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2198: {
1.5 noro 2199: int i,j,td,len,max;
1.1 noro 2200: NODE s,s0,f0;
1.61 noro 2201: UINT *d;
1.13 noro 2202: RHist r;
1.69 noro 2203: NDV *w;
1.61 noro 2204: NDV a,am;
1.11 noro 2205:
2206: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2207:
1.69 noro 2208: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2209: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2210: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.128 noro 2211: if ( !dont_sort )
2212: qsort(w,nd_psn,sizeof(NDV),
2213: (int (*)(const void *,const void *))ndv_compare);
1.69 noro 2214: nd_pslen = 2*nd_psn;
1.11 noro 2215: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2216: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2217: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2218: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2219:
1.11 noro 2220: if ( !nd_red )
1.13 noro 2221: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2222: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2223: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2224: if ( trace ) {
1.69 noro 2225: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2226: ndv_removecont(0,a);
2227: am = nd_ps[i] = ndv_dup(mod,a);
2228: ndv_mod(mod,am);
2229: ndv_removecont(mod,am);
2230: } else {
1.69 noro 2231: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2232: ndv_removecont(mod,a);
1.39 noro 2233: }
1.61 noro 2234: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2235: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2236: nd_psh[i] = r;
1.77 noro 2237: if ( nd_demand ) {
2238: if ( trace ) {
2239: ndv_save(nd_ps_trace[i],i);
2240: nd_ps_trace[i] = 0;
2241: } else {
2242: ndv_save(nd_ps[i],i);
2243: nd_ps[i] = 0;
2244: }
2245: }
1.20 noro 2246: }
2247: }
2248:
1.119 noro 2249: struct order_spec *append_block(struct order_spec *spec,
2250: int nv,int nalg,int ord);
2251:
1.121 noro 2252: extern VECT current_dl_weight_vector_obj;
2253: static VECT prev_weight_vector_obj;
2254:
1.120 noro 2255: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2256: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2257: {
1.121 noro 2258: NODE alist,t,s,r0,r,arg;
1.120 noro 2259: VL tv;
2260: P poly;
2261: DP d;
2262: Alg alpha,dp;
2263: DAlg inv,da,hc;
2264: MP m;
1.121 noro 2265: int i,nvar,nalg,n;
1.120 noro 2266: NumberField nf;
2267: LIST f1,f2;
2268: struct order_spec *current_spec;
1.121 noro 2269: VECT obj,obj0;
2270: Obj tmp;
1.120 noro 2271:
2272: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2273: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2274:
2275: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2276: NEXTNODE(alist,t); MKV(tv->v,poly);
2277: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2278: tv->v = tv->v->priv;
2279: }
2280: NEXT(t) = 0;
2281:
2282: /* simplification, makeing polynomials monic */
2283: setfield_dalg(alist);
2284: obj_algtodalg(f,&f1);
2285: for ( t = BDY(f); t; t = NEXT(t) ) {
2286: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2287: hc = (DAlg)BDY(d)->c;
2288: if ( NID(hc) == N_DA ) {
2289: invdalg(hc,&inv);
2290: for ( m = BDY(d); m; m = NEXT(m) ) {
2291: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2292: }
2293: }
2294: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2295: }
2296: obj_dalgtoalg(f1,&f);
2297:
2298: /* append alg vars to the var list */
2299: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2300: NEXT(tv) = av;
2301:
2302: /* append a block to ord */
2303: *ord1p = append_block(ord,nvar,nalg,2);
2304:
2305: /* create generator list */
2306: nf = get_numberfield();
2307: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2308: MKAlg(nf->defpoly[i],dp);
2309: MKNODE(s,dp,t); t = s;
2310: }
2311: MKLIST(f1,t);
2312: *alistp = alist;
2313: algobjtorat(f1,f1p);
1.121 noro 2314:
2315: /* creating a new weight vector */
2316: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2317: n = nvar+nalg+1;
2318: MKVECT(obj,n);
2319: if ( obj0 && obj0->len == nvar )
2320: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2321: else
2322: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2323: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2324: BDY(obj)[n-1] = (pointer)ONE;
2325: arg = mknode(1,obj);
2326: Pdp_set_weight(arg,&tmp);
2327: }
2328:
2329: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2330: {
2331: NODE s,t,u0,u;
2332: P p;
2333: VL tv;
2334: Obj obj,tmp;
2335: NODE arg;
2336:
2337: u0 = 0;
2338: for ( t = r; t; t = NEXT(t) ) {
2339: p = (P)BDY(t);
2340: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2341: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2342: }
2343: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2344: NEXTNODE(u0,u);
2345: BDY(u) = (pointer)p;
2346: }
2347: }
2348: arg = mknode(1,prev_weight_vector_obj);
2349: Pdp_set_weight(arg,&tmp);
2350:
2351: return u0;
1.120 noro 2352: }
2353:
1.63 noro 2354: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2355: {
1.119 noro 2356: VL tv,fv,vv,vc,av;
2357: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2358: int e,max,nvar,i;
1.69 noro 2359: NDV b;
1.119 noro 2360: int ishomo,nalg;
2361: Alg alpha,dp;
2362: P p;
2363: LIST f1,f2;
2364: Obj obj;
2365: NumberField nf;
1.120 noro 2366: struct order_spec *ord1;
1.1 noro 2367:
1.78 noro 2368: if ( !m && Demand ) nd_demand = 1;
2369: else nd_demand = 0;
2370:
1.103 noro 2371: #if 0
1.74 noro 2372: ndv_alloc = 0;
1.103 noro 2373: #endif
1.1 noro 2374: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2375: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2376: switch ( ord->id ) {
2377: case 1:
2378: if ( ord->nv != nvar )
2379: error("nd_{gr,f4} : invalid order specification");
2380: break;
2381: default:
2382: break;
2383: }
1.119 noro 2384: nd_nalg = 0;
1.123 noro 2385: av = 0;
1.119 noro 2386: if ( !m ) {
2387: get_algtree((Obj)f,&av);
2388: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2389: nd_ntrans = nvar;
1.119 noro 2390: nd_nalg = nalg;
2391: /* #i -> t#i */
2392: if ( nalg ) {
1.120 noro 2393: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2394: ord = ord1;
2395: f = f1;
1.119 noro 2396: }
1.120 noro 2397: nvar += nalg;
1.119 noro 2398: }
1.32 noro 2399: nd_init_ord(ord);
1.61 noro 2400: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2401: for ( tv = vv; tv; tv = NEXT(tv) ) {
2402: e = getdeg(tv->v,(P)BDY(t));
2403: max = MAX(e,max);
2404: }
2405: nd_setup_parameters(nvar,max);
1.77 noro 2406: ishomo = 1;
1.1 noro 2407: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2408: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2409: if ( ishomo )
2410: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2411: if ( m ) ndv_mod(m,b);
1.61 noro 2412: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2413: }
2414: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2415: ndv_setup(m,0,fd0,0);
1.77 noro 2416: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2417: nd_demand = 0;
1.61 noro 2418: x = ndv_reducebase(x);
2419: x = ndv_reduceall(m,x);
1.23 noro 2420: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2421: NEXTNODE(r0,r);
1.61 noro 2422: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2423: }
2424: if ( r0 ) NEXT(r) = 0;
1.121 noro 2425: if ( nalg )
2426: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2427: MKLIST(*rp,r0);
1.103 noro 2428: #if 0
1.74 noro 2429: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2430: #endif
1.127 noro 2431: }
2432:
2433: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2434: {
2435: VL tv,fv,vv,vc,av;
2436: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2437: int e,max,nvar,i;
2438: NDV b;
2439: int ishomo,nalg;
2440: Alg alpha,dp;
2441: P p;
2442: LIST f1,f2;
2443: Obj obj;
2444: NumberField nf;
2445: struct order_spec *ord1;
2446:
2447: get_vars((Obj)f,&fv); pltovl(v,&vv);
2448: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2449: switch ( ord->id ) {
2450: case 1:
2451: if ( ord->nv != nvar )
2452: error("nd_check : invalid order specification");
2453: break;
2454: default:
2455: break;
2456: }
2457: nd_nalg = 0;
2458: av = 0;
2459: if ( !m ) {
2460: get_algtree((Obj)f,&av);
2461: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2462: nd_ntrans = nvar;
2463: nd_nalg = nalg;
2464: /* #i -> t#i */
2465: if ( nalg ) {
2466: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2467: ord = ord1;
2468: f = f1;
2469: }
2470: nvar += nalg;
2471: }
2472: nd_init_ord(ord);
2473: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2474: for ( tv = vv; tv; tv = NEXT(tv) ) {
2475: e = getdeg(tv->v,(P)BDY(t));
2476: max = MAX(e,max);
2477: }
2478: nd_setup_parameters(nvar,max);
2479: ishomo = 1;
2480: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2481: b = (pointer)ptondv(CO,vv,(P)BDY(t));
2482: if ( ishomo )
2483: ishomo = ishomo && ndv_ishomo(b);
2484: if ( m ) ndv_mod(m,b);
2485: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2486: }
2487: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2488: ndv_setup(m,0,fd0,0);
1.127 noro 2489: for ( x = 0, i = 0; i < nd_psn; i++ )
2490: x = update_base(x,i);
2491: if ( do_check ) {
2492: x = nd_gb(m,ishomo,1);
2493: if ( !x ) {
2494: *rp = 0;
2495: return;
2496: }
2497: } else {
2498: for ( t = x; t; t = NEXT(t) )
2499: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2500: }
2501: x = ndv_reducebase(x);
2502: x = ndv_reduceall(m,x);
2503: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2504: NEXTNODE(r0,r);
2505: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2506: }
2507: if ( r0 ) NEXT(r) = 0;
2508: if ( nalg )
2509: r0 = postprocess_algcoef(av,alist,r0);
2510: MKLIST(*rp,r0);
1.20 noro 2511: }
2512:
1.133 noro 2513: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2514: {
1.119 noro 2515: VL tv,fv,vv,vc,av;
2516: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2517: int m,nocheck,nvar,mindex,e,max;
2518: NDV c;
2519: NMV a;
1.27 noro 2520: P p;
1.61 noro 2521: EPOS oepos;
1.119 noro 2522: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2523: Alg alpha,dp;
2524: P poly;
2525: LIST f1,f2;
2526: Obj obj;
2527: NumberField nf;
1.120 noro 2528: struct order_spec *ord1;
1.20 noro 2529:
2530: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2531: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2532: switch ( ord->id ) {
2533: case 1:
2534: if ( ord->nv != nvar )
2535: error("nd_gr_trace : invalid order specification");
2536: break;
2537: default:
2538: break;
2539: }
1.119 noro 2540:
2541: get_algtree((Obj)f,&av);
2542: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2543: nd_ntrans = nvar;
1.119 noro 2544: nd_nalg = nalg;
2545: /* #i -> t#i */
2546: if ( nalg ) {
1.120 noro 2547: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2548: ord = ord1;
2549: f = f1;
1.119 noro 2550: }
1.120 noro 2551: nvar += nalg;
1.119 noro 2552:
1.52 noro 2553: nocheck = 0;
2554: mindex = 0;
2555:
1.78 noro 2556: if ( Demand ) nd_demand = 1;
2557: else nd_demand = 0;
2558:
1.52 noro 2559: /* setup modulus */
2560: if ( trace < 0 ) {
2561: trace = -trace;
2562: nocheck = 1;
2563: }
2564: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2565: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2566: for ( tv = vv; tv; tv = NEXT(tv) ) {
2567: e = getdeg(tv->v,(P)BDY(t));
2568: max = MAX(e,max);
1.23 noro 2569: }
1.61 noro 2570: nd_init_ord(ord);
2571: nd_setup_parameters(nvar,max);
2572: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2573: ishomo = 1;
1.61 noro 2574: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2575: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2576: if ( ishomo )
2577: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2578: if ( c ) {
2579: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2580: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2581: }
1.61 noro 2582: }
2583: if ( in0 ) NEXT(in) = 0;
2584: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2585: if ( !ishomo && homo ) {
1.122 noro 2586: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2587: c = (NDV)BDY(t); len = LEN(c);
2588: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2589: wmax = MAX(TD(DL(a)),wmax);
2590: }
2591: homogenize_order(ord,nvar,&ord1);
1.89 noro 2592: nd_init_ord(ord1);
1.61 noro 2593: nd_setup_parameters(nvar+1,wmax);
2594: for ( t = fd0; t; t = NEXT(t) )
2595: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2596: }
1.52 noro 2597: while ( 1 ) {
1.77 noro 2598: if ( Demand )
2599: nd_demand = 1;
1.128 noro 2600: ndv_setup(m,1,fd0,0);
1.133 noro 2601: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
1.52 noro 2602: if ( !cand ) {
2603: /* failure */
1.61 noro 2604: if ( trace > 1 ) { *rp = 0; return; }
2605: else m = get_lprime(++mindex);
1.52 noro 2606: continue;
2607: }
1.82 noro 2608: if ( !ishomo && homo ) {
1.27 noro 2609: /* dehomogenization */
1.61 noro 2610: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2611: nd_init_ord(ord);
1.61 noro 2612: nd_setup_parameters(nvar,0);
1.27 noro 2613: }
1.77 noro 2614: nd_demand = 0;
1.61 noro 2615: cand = ndv_reducebase(cand);
2616: cand = ndv_reduceall(0,cand);
1.78 noro 2617: cbpe = nd_bpe;
1.61 noro 2618: if ( nocheck )
2619: break;
2620: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2621: /* success */
2622: break;
2623: else if ( trace > 1 ) {
2624: /* failure */
2625: *rp = 0; return;
1.61 noro 2626: } else {
1.52 noro 2627: /* try the next modulus */
2628: m = get_lprime(++mindex);
1.61 noro 2629: /* reset the parameters */
1.82 noro 2630: if ( !ishomo && homo ) {
1.89 noro 2631: nd_init_ord(ord1);
1.61 noro 2632: nd_setup_parameters(nvar+1,wmax);
2633: } else {
2634: nd_init_ord(ord);
2635: nd_setup_parameters(nvar,max);
2636: }
2637: }
1.52 noro 2638: }
1.27 noro 2639: /* dp->p */
1.61 noro 2640: nd_bpe = cbpe;
1.73 noro 2641: nd_setup_parameters(nd_nvar,0);
1.121 noro 2642: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2643: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2644: if ( nalg )
2645: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2646: MKLIST(*rp,cand);
1.1 noro 2647: }
2648:
1.61 noro 2649: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2650: {
1.61 noro 2651: UINT *d;
1.57 noro 2652: int i,j,l,s,ord_l;
1.43 noro 2653: struct order_pair *op;
1.1 noro 2654:
2655: d = dl->d;
1.41 noro 2656: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2657: if ( nd_blockmask ) {
2658: l = nd_blockmask->n;
2659: op = nd_blockmask->order_pair;
2660: for ( j = 0, s = 0; j < l; j++ ) {
2661: ord_l = op[j].length;
1.57 noro 2662: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2663: }
2664: TD(r) = ndl_weight(r);
1.61 noro 2665: ndl_weight_mask(r);
1.43 noro 2666: } else {
1.56 noro 2667: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2668: TD(r) = ndl_weight(r);
2669: }
1.1 noro 2670: }
2671:
1.61 noro 2672: DL ndltodl(int n,UINT *ndl)
1.1 noro 2673: {
2674: DL dl;
2675: int *d;
1.57 noro 2676: int i,j,l,s,ord_l;
1.43 noro 2677: struct order_pair *op;
1.1 noro 2678:
2679: NEWDL(dl,n);
1.34 noro 2680: dl->td = TD(ndl);
1.1 noro 2681: d = dl->d;
1.43 noro 2682: if ( nd_blockmask ) {
2683: l = nd_blockmask->n;
2684: op = nd_blockmask->order_pair;
2685: for ( j = 0, s = 0; j < l; j++ ) {
2686: ord_l = op[j].length;
1.57 noro 2687: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2688: }
2689: } else {
1.56 noro 2690: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2691: }
1.1 noro 2692: return dl;
2693: }
2694:
1.61 noro 2695: void ndl_print(UINT *dl)
1.1 noro 2696: {
2697: int n;
1.57 noro 2698: int i,j,l,ord_l,s,s0;
1.43 noro 2699: struct order_pair *op;
1.1 noro 2700:
2701: n = nd_nvar;
2702: printf("<<");
1.43 noro 2703: if ( nd_blockmask ) {
2704: l = nd_blockmask->n;
2705: op = nd_blockmask->order_pair;
2706: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2707: ord_l = op[j].length;
1.57 noro 2708: for ( i = 0; i < ord_l; i++, s++ )
2709: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2710: }
2711: } else {
1.56 noro 2712: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2713: }
1.1 noro 2714: printf(">>");
2715: }
2716:
2717: void nd_print(ND p)
2718: {
2719: NM m;
2720:
2721: if ( !p )
2722: printf("0\n");
2723: else {
2724: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2725: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2726: else printf("+%d*",CM(m));
1.14 noro 2727: ndl_print(DL(m));
1.1 noro 2728: }
2729: printf("\n");
2730: }
2731: }
2732:
1.113 noro 2733: void nd_print_q(ND p)
1.16 noro 2734: {
2735: NM m;
2736:
2737: if ( !p )
2738: printf("0\n");
2739: else {
2740: for ( m = BDY(p); m; m = NEXT(m) ) {
2741: printf("+");
1.113 noro 2742: printexpr(CO,(Obj)CQ(m));
1.16 noro 2743: printf("*");
2744: ndl_print(DL(m));
2745: }
2746: printf("\n");
2747: }
2748: }
2749:
1.1 noro 2750: void ndp_print(ND_pairs d)
2751: {
2752: ND_pairs t;
2753:
1.34 noro 2754: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2755: printf("\n");
2756: }
2757:
1.20 noro 2758: void nd_removecont(int mod,ND p)
1.16 noro 2759: {
2760: int i,n;
1.113 noro 2761: Q *w;
2762: Q dvr,t;
1.16 noro 2763: NM m;
1.21 noro 2764: struct oVECT v;
1.113 noro 2765: N q,r;
1.16 noro 2766:
1.71 noro 2767: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2768: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2769: else {
2770: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2771: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2772: v.len = n;
2773: v.body = (pointer *)w;
1.113 noro 2774: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2775: removecont_array(w,n);
1.113 noro 2776: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2777: }
2778: }
2779:
1.21 noro 2780: void nd_removecont2(ND p1,ND p2)
2781: {
2782: int i,n1,n2,n;
1.113 noro 2783: Q *w;
2784: Q dvr,t;
1.21 noro 2785: NM m;
2786: struct oVECT v;
1.113 noro 2787: N q,r;
1.21 noro 2788:
2789: if ( !p1 ) {
2790: nd_removecont(0,p2); return;
2791: } else if ( !p2 ) {
2792: nd_removecont(0,p1); return;
2793: }
2794: n1 = nd_length(p1);
2795: n2 = nd_length(p2);
2796: n = n1+n2;
1.113 noro 2797: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2798: v.len = n;
2799: v.body = (pointer *)w;
1.113 noro 2800: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2801: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2802: removecont_array(w,n);
1.113 noro 2803: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2804: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2805: }
2806:
1.20 noro 2807: void ndv_removecont(int mod,NDV p)
1.16 noro 2808: {
2809: int i,len;
1.113 noro 2810: Q *w;
2811: Q dvr,t;
1.16 noro 2812: NMV m;
2813:
1.71 noro 2814: if ( mod == -1 )
2815: ndv_mul_c(mod,p,_invsf(HCM(p)));
2816: else if ( mod )
1.20 noro 2817: ndv_mul_c(mod,p,invm(HCM(p),mod));
2818: else {
2819: len = p->len;
1.113 noro 2820: w = (Q *)ALLOCA(len*sizeof(Q));
2821: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2822: sortbynm(w,len);
2823: qltozl(w,len,&dvr);
2824: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2825: divq(CQ(m),dvr,&t); CQ(m) = t;
2826: }
1.16 noro 2827: }
1.21 noro 2828: }
2829:
1.61 noro 2830: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2831: {
2832: int len,i,max;
2833: NMV m,mr0,mr,t;
2834:
2835: len = p->len;
2836: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2837: max = MAX(max,TD(DL(m)));
2838: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2839: m = (NMV)((char *)mr0+(len-1)*oadv);
2840: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2841: t = (NMV)ALLOCA(nmv_adv);
2842: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2843: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2844: CQ(mr) = CQ(m);
1.61 noro 2845: ndl_copy(DL(t),DL(mr));
2846: }
2847: NV(p)++;
2848: BDY(p) = mr0;
2849: }
2850:
1.45 noro 2851: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2852: {
1.45 noro 2853: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2854: Q *w;
2855: Q dvr,t;
1.23 noro 2856: NMV m,r;
2857:
2858: len = p->len;
2859: newnvar = nd_nvar-1;
1.48 noro 2860: newexporigin = nd_get_exporigin(ord);
1.45 noro 2861: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2862: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2863: ndl_dehomogenize(DL(m));
1.23 noro 2864: if ( newwpd != nd_wpd ) {
1.90 noro 2865: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2866: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2867: CQ(r) = CQ(m);
1.45 noro 2868: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2869: adj = nd_exporigin-newexporigin;
2870: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2871: }
2872: }
2873: NV(p)--;
2874: }
2875:
1.113 noro 2876: void removecont_array(Q *c,int n)
1.21 noro 2877: {
2878: struct oVECT v;
1.113 noro 2879: Q d0,d1,a,u,u1,gcd;
1.110 noro 2880: int i,j;
1.113 noro 2881: N qn,rn,gn;
2882: Q *q,*r;
1.21 noro 2883:
1.113 noro 2884: q = (Q *)ALLOCA(n*sizeof(Q));
2885: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2886: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2887: igcdv_estimate(&v,&d0);
2888: for ( i = 0; i < n; i++ ) {
2889: divn(NM(c[i]),NM(d0),&qn,&rn);
2890: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2891: NTOQ(rn,SGN(c[i]),r[i]);
2892: }
1.34 noro 2893: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2894: if ( i < n ) {
2895: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2896: igcdv(&v,&d1);
2897: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2898: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2899: for ( i = 0; i < n; i++ ) {
1.113 noro 2900: mulq(a,q[i],&u);
2901: if ( r[i] ) {
2902: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2903: addq(u,u1,&q[i]);
2904: } else
2905: q[i] = u;
1.21 noro 2906: }
2907: }
1.34 noro 2908: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2909: }
2910:
1.19 noro 2911: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2912: {
2913: NM m;
2914: int c,c1;
2915:
1.34 noro 2916: if ( !p ) return;
1.115 noro 2917: if ( mul == 1 ) return;
1.71 noro 2918: if ( mod == -1 )
2919: for ( m = BDY(p); m; m = NEXT(m) )
2920: CM(m) = _mulsf(CM(m),mul);
2921: else
2922: for ( m = BDY(p); m; m = NEXT(m) ) {
2923: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2924: }
1.1 noro 2925: }
2926:
1.113 noro 2927: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2928: {
2929: NM m;
1.113 noro 2930: Q c;
1.16 noro 2931:
1.113 noro 2932: if ( !p ) return;
1.115 noro 2933: if ( UNIQ(mul) ) return;
1.16 noro 2934: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2935: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2936: }
2937: }
2938:
1.61 noro 2939: void nd_mul_c_p(VL vl,ND p,P mul)
2940: {
2941: NM m;
2942: P c;
2943:
2944: if ( !p ) return;
2945: for ( m = BDY(p); m; m = NEXT(m) ) {
2946: mulp(vl,CP(m),mul,&c); CP(m) = c;
2947: }
2948: }
2949:
1.1 noro 2950: void nd_free(ND p)
2951: {
2952: NM t,s;
2953:
1.34 noro 2954: if ( !p ) return;
1.1 noro 2955: t = BDY(p);
2956: while ( t ) {
2957: s = NEXT(t);
2958: FREENM(t);
2959: t = s;
2960: }
2961: FREEND(p);
2962: }
2963:
1.23 noro 2964: void ndv_free(NDV p)
2965: {
2966: GC_free(BDY(p));
2967: }
2968:
1.61 noro 2969: void nd_append_red(UINT *d,int i)
1.1 noro 2970: {
1.13 noro 2971: RHist m,m0;
1.1 noro 2972: int h;
2973:
1.13 noro 2974: NEWRHist(m);
1.34 noro 2975: h = ndl_hash_value(d);
1.13 noro 2976: m->index = i;
1.14 noro 2977: ndl_copy(d,DL(m));
1.1 noro 2978: NEXT(m) = nd_red[h];
2979: nd_red[h] = m;
2980: }
2981:
1.61 noro 2982: UINT *ndv_compute_bound(NDV p)
1.1 noro 2983: {
1.61 noro 2984: UINT *d1,*d2,*t;
2985: UINT u;
1.57 noro 2986: int i,j,k,l,len,ind;
1.45 noro 2987: NMV m;
1.1 noro 2988:
2989: if ( !p )
2990: return 0;
1.61 noro 2991: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2992: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2993: len = LEN(p);
2994: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2995: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2996: ndl_lcm(DL(m),d1,d2);
1.1 noro 2997: t = d1; d1 = d2; d2 = t;
2998: }
1.12 noro 2999: l = nd_nvar+31;
1.61 noro 3000: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3001: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3002: u = d1[i];
3003: k = (nd_epw-1)*nd_bpe;
3004: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3005: t[ind] = (u>>k)&nd_mask0;
3006: }
3007: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3008: return t;
3009: }
3010:
1.99 noro 3011: UINT *nd_compute_bound(ND p)
3012: {
3013: UINT *d1,*d2,*t;
3014: UINT u;
3015: int i,j,k,l,len,ind;
3016: NM m;
3017:
3018: if ( !p )
3019: return 0;
3020: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3021: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3022: len = LEN(p);
3023: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3024: for ( m = NEXT(m); m; m = NEXT(m) ) {
3025: ndl_lcm(DL(m),d1,d2);
3026: t = d1; d1 = d2; d2 = t;
3027: }
3028: l = nd_nvar+31;
3029: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3030: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3031: u = d1[i];
3032: k = (nd_epw-1)*nd_bpe;
3033: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3034: t[ind] = (u>>k)&nd_mask0;
3035: }
3036: for ( ; ind < l; ind++ ) t[ind] = 0;
3037: return t;
3038: }
3039:
1.48 noro 3040: int nd_get_exporigin(struct order_spec *ord)
3041: {
1.51 noro 3042: switch ( ord->id ) {
1.96 noro 3043: case 0: case 2:
1.48 noro 3044: return 1;
1.41 noro 3045: case 1:
3046: /* block order */
1.43 noro 3047: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3048: return ord->ord.block.length+1;
1.96 noro 3049: case 3:
3050: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3051: }
1.48 noro 3052: }
3053:
1.61 noro 3054: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3055: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3056: struct order_pair *op;
1.48 noro 3057:
1.73 noro 3058: nd_nvar = nvar;
1.79 noro 3059: if ( max ) {
1.83 noro 3060: /* XXX */
3061: if ( do_weyl ) nd_bpe = 32;
3062: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3063: else if ( max < 4 ) nd_bpe = 2;
3064: else if ( max < 8 ) nd_bpe = 3;
3065: else if ( max < 16 ) nd_bpe = 4;
3066: else if ( max < 32 ) nd_bpe = 5;
3067: else if ( max < 64 ) nd_bpe = 6;
3068: else if ( max < 256 ) nd_bpe = 8;
3069: else if ( max < 1024 ) nd_bpe = 10;
3070: else if ( max < 65536 ) nd_bpe = 16;
3071: else nd_bpe = 32;
3072: }
1.61 noro 3073: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3074: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3075: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3076: wpd = nd_exporigin+elen;
3077: if ( wpd != nd_wpd ) {
3078: nd_free_private_storage();
3079: nd_wpd = wpd;
3080: }
1.1 noro 3081: if ( nd_bpe < 32 ) {
3082: nd_mask0 = (1<<nd_bpe)-1;
3083: } else {
3084: nd_mask0 = 0xffffffff;
3085: }
3086: bzero(nd_mask,sizeof(nd_mask));
3087: nd_mask1 = 0;
3088: for ( i = 0; i < nd_epw; i++ ) {
3089: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3090: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3091: }
1.90 noro 3092: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3093: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3094: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3095: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3096: }
3097:
1.103 noro 3098: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3099: {
1.37 noro 3100: int i,obpe,oadv,h;
1.71 noro 3101: static NM prev_nm_free_list;
3102: static ND_pairs prev_ndp_free_list;
1.13 noro 3103: RHist mr0,mr;
3104: RHist r;
1.37 noro 3105: RHist *old_red;
1.71 noro 3106: ND_pairs s0,s,t;
1.43 noro 3107: EPOS oepos;
1.15 noro 3108:
1.1 noro 3109: obpe = nd_bpe;
1.11 noro 3110: oadv = nmv_adv;
1.43 noro 3111: oepos = nd_epos;
1.63 noro 3112: if ( obpe < 2 ) nd_bpe = 2;
3113: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3114: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3115: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3116: else if ( obpe < 6 ) nd_bpe = 6;
3117: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3118: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3119: else if ( obpe < 16 ) nd_bpe = 16;
3120: else if ( obpe < 32 ) nd_bpe = 32;
3121: else error("nd_reconstruct : exponent too large");
1.5 noro 3122:
1.73 noro 3123: nd_setup_parameters(nd_nvar,0);
1.1 noro 3124: prev_nm_free_list = _nm_free_list;
3125: prev_ndp_free_list = _ndp_free_list;
3126: _nm_free_list = 0;
3127: _ndp_free_list = 0;
1.53 noro 3128: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3129: if ( trace )
3130: for ( i = nd_psn-1; i >= 0; i-- )
3131: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3132: s0 = 0;
3133: for ( t = d; t; t = NEXT(t) ) {
3134: NEXTND_pairs(s0,s);
3135: s->i1 = t->i1;
3136: s->i2 = t->i2;
1.14 noro 3137: SG(s) = SG(t);
1.61 noro 3138: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3139: }
1.37 noro 3140:
3141: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3142: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3143: old_red[i] = nd_red[i];
3144: nd_red[i] = 0;
3145: }
3146: for ( i = 0; i < REDTAB_LEN; i++ )
3147: for ( r = old_red[i]; r; r = NEXT(r) ) {
3148: NEWRHist(mr);
1.13 noro 3149: mr->index = r->index;
1.20 noro 3150: SG(mr) = SG(r);
1.61 noro 3151: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3152: h = ndl_hash_value(DL(mr));
3153: NEXT(mr) = nd_red[h];
3154: nd_red[h] = mr;
1.6 noro 3155: }
1.37 noro 3156: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3157: old_red = 0;
1.11 noro 3158: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3159: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3160: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3161: nd_psh[i] = r;
1.11 noro 3162: }
1.1 noro 3163: if ( s0 ) NEXT(s) = 0;
3164: prev_nm_free_list = 0;
3165: prev_ndp_free_list = 0;
1.71 noro 3166: #if 0
1.1 noro 3167: GC_gcollect();
1.71 noro 3168: #endif
1.1 noro 3169: return s0;
3170: }
3171:
1.61 noro 3172: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3173: {
1.57 noro 3174: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3175: struct order_pair *op;
1.1 noro 3176:
3177: n = nd_nvar;
1.61 noro 3178: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3179: omask0 = (1<<obpe)-1;
1.34 noro 3180: TD(r) = TD(d);
1.41 noro 3181: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3182: if ( nd_blockmask ) {
3183: l = nd_blockmask->n;
3184: op = nd_blockmask->order_pair;
3185: for ( i = 1; i < nd_exporigin; i++ )
3186: r[i] = d[i];
3187: for ( j = 0, s = 0; j < l; j++ ) {
3188: ord_l = op[j].length;
1.57 noro 3189: for ( i = 0; i < ord_l; i++, s++ ) {
3190: ei = GET_EXP_OLD(d,s);
3191: PUT_EXP(r,s,ei);
3192: }
1.1 noro 3193: }
1.43 noro 3194: } else {
1.56 noro 3195: for ( i = 0; i < n; i++ ) {
3196: ei = GET_EXP_OLD(d,i);
3197: PUT_EXP(r,i,ei);
3198: }
1.1 noro 3199: }
3200: }
1.3 noro 3201:
1.6 noro 3202: ND nd_copy(ND p)
3203: {
3204: NM m,mr,mr0;
1.41 noro 3205: int c,n;
1.6 noro 3206: ND r;
3207:
3208: if ( !p )
3209: return 0;
3210: else {
3211: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3212: NEXTNM(mr0,mr);
1.14 noro 3213: CM(mr) = CM(m);
3214: ndl_copy(DL(m),DL(mr));
1.6 noro 3215: }
3216: NEXT(mr) = 0;
1.31 noro 3217: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3218: SG(r) = SG(p);
1.6 noro 3219: return r;
3220: }
3221: }
3222:
1.53 noro 3223: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3224: {
3225: NM m;
3226: NDV p1,p2;
3227: ND t1,t2;
1.61 noro 3228: UINT *lcm;
1.31 noro 3229: int td;
1.11 noro 3230:
1.77 noro 3231: if ( !mod && nd_demand ) {
3232: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3233: } else {
1.77 noro 3234: if ( trace ) {
3235: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3236: } else {
3237: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3238: }
1.20 noro 3239: }
1.34 noro 3240: lcm = LCM(p);
1.11 noro 3241: NEWNM(m);
1.113 noro 3242: CQ(m) = HCQ(p2);
1.34 noro 3243: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3244: if ( ndl_check_bound2(p->i1,DL(m)) )
3245: return 0;
1.55 noro 3246: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3247: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3248: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3249: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3250: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3251: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3252: nd_free(t1);
3253: return 0;
3254: }
1.55 noro 3255: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3256: *rp = nd_add(mod,t1,t2);
1.11 noro 3257: FREENM(m);
3258: return 1;
3259: }
3260:
1.19 noro 3261: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3262: {
3263: NMV m;
3264: int c,c1,len,i;
3265:
1.34 noro 3266: if ( !p ) return;
1.14 noro 3267: len = LEN(p);
1.71 noro 3268: if ( mod == -1 )
3269: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3270: CM(m) = _mulsf(CM(m),mul);
3271: else
3272: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3273: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3274: }
1.11 noro 3275: }
3276:
1.113 noro 3277: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3278: {
3279: NMV m;
1.113 noro 3280: Q c;
1.16 noro 3281: int len,i;
3282:
1.34 noro 3283: if ( !p ) return;
1.16 noro 3284: len = LEN(p);
3285: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3286: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3287: }
3288: }
3289:
1.55 noro 3290: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3291: int n2,i,j,l,n,tlen;
1.61 noro 3292: UINT *d0;
1.55 noro 3293: NM *tab,*psum;
3294: ND s,r;
3295: NM t;
3296: NMV m1;
3297:
3298: if ( !p ) return 0;
3299: n = NV(p); n2 = n>>1;
3300: d0 = DL(m0);
3301: l = LEN(p);
3302: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3303: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3304: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3305: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3306: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3307: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3308: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3309: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3310: for ( j = 0; j < tlen; j++ ) {
3311: if ( tab[j] ) {
3312: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3313: }
3314: }
3315: }
3316: for ( i = tlen-1, r = 0; i >= 0; i-- )
3317: if ( psum[i] ) {
3318: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3319: MKND(n,psum[i],j,s);
3320: r = nd_add(mod,r,s);
3321: }
1.56 noro 3322: if ( r ) SG(r) = SG(p)+TD(d0);
3323: return r;
1.55 noro 3324: }
3325:
1.56 noro 3326: /* product of monomials */
3327: /* XXX block order is not handled correctly */
3328:
1.55 noro 3329: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3330: {
1.56 noro 3331: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3332: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3333: Q *ctab_q;
3334: Q q,q1;
1.61 noro 3335: UINT c0,c1,c;
1.55 noro 3336: NM *p;
3337: NM m,t;
3338:
3339: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3340: if ( !m0 || !m1 ) return;
3341: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3342: NEWNM(m); d = DL(m);
1.56 noro 3343: if ( mod ) {
3344: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3345: } else
1.113 noro 3346: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3347: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3348: homo = n&1 ? 1 : 0;
3349: if ( homo ) {
3350: /* offset of h-degree */
3351: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3352: PUT_EXP(DL(m),n-1,h);
3353: TD(DL(m)) = h;
1.61 noro 3354: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3355: }
3356: tab[0] = m;
3357: NEWNM(m); d = DL(m);
1.57 noro 3358: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3359: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3360: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3361: /* xi^a*(Di^k*xi^l)*Di^b */
3362: a += l; b += k;
1.56 noro 3363: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3364: if ( !k || !l ) {
3365: for ( j = 0; j < curlen; j++ )
1.56 noro 3366: if ( t = tab[j] ) {
3367: dt = DL(t);
3368: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3369: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3370: }
3371: curlen *= k+1;
3372: continue;
3373: }
3374: min = MIN(k,l);
1.56 noro 3375: if ( mod ) {
1.61 noro 3376: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3377: mkwcm(k,l,mod,ctab);
3378: } else {
1.113 noro 3379: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3380: mkwc(k,l,ctab_q);
1.56 noro 3381: }
1.57 noro 3382: for ( j = min; j >= 0; j-- ) {
1.56 noro 3383: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3384: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3385: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3386: if ( homo ) {
3387: TD(d) = s;
1.56 noro 3388: PUT_EXP(d,n-1,s-h);
1.55 noro 3389: } else TD(d) = h;
1.61 noro 3390: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3391: if ( mod ) c = ctab[j];
1.113 noro 3392: else q = ctab_q[j];
1.57 noro 3393: p = tab+curlen*j;
3394: if ( j == 0 ) {
3395: for ( u = 0; u < curlen; u++, p++ ) {
3396: if ( tab[u] ) {
3397: ndl_addto(DL(tab[u]),d);
3398: if ( mod ) {
3399: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3400: } else {
1.113 noro 3401: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3402: }
3403: }
1.56 noro 3404: }
1.57 noro 3405: } else {
3406: for ( u = 0; u < curlen; u++, p++ ) {
3407: if ( tab[u] ) {
3408: NEWNM(t);
3409: ndl_add(DL(tab[u]),d,DL(t));
3410: if ( mod ) {
3411: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3412: } else
1.113 noro 3413: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3414: *p = t;
3415: }
1.55 noro 3416: }
3417: }
3418: }
3419: curlen *= k+1;
3420: }
3421: FREENM(m);
3422: }
3423:
1.63 noro 3424: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3425: {
3426: NM mr,mr0;
3427: NMV m;
3428: UINT *d,*dt,*dm;
3429: int c,n,td,i,c1,c2,len;
1.113 noro 3430: Q q;
1.63 noro 3431: ND r;
3432:
3433: if ( !p ) return 0;
3434: else {
3435: n = NV(p); m = BDY(p);
3436: d = DL(m0);
3437: len = LEN(p);
3438: mr0 = 0;
3439: td = TD(d);
3440: c = CM(m0);
3441: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3442: NEXTNM(mr0,mr);
3443: CM(mr) = 1;
3444: ndl_add(DL(m),d,DL(mr));
3445: }
3446: NEXT(mr) = 0;
3447: MKND(NV(p),mr0,len,r);
3448: SG(r) = SG(p) + TD(d);
3449: return r;
3450: }
3451: }
3452:
1.55 noro 3453: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3454: {
3455: NM mr,mr0;
3456: NMV m;
1.61 noro 3457: UINT *d,*dt,*dm;
1.9 noro 3458: int c,n,td,i,c1,c2,len;
1.113 noro 3459: Q q;
1.9 noro 3460: ND r;
3461:
1.34 noro 3462: if ( !p ) return 0;
1.55 noro 3463: else if ( do_weyl )
1.71 noro 3464: if ( mod == -1 )
3465: error("ndv_mul_nm : not implemented (weyl)");
3466: else
3467: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3468: else {
3469: n = NV(p); m = BDY(p);
1.34 noro 3470: d = DL(m0);
1.14 noro 3471: len = LEN(p);
1.9 noro 3472: mr0 = 0;
1.34 noro 3473: td = TD(d);
1.71 noro 3474: if ( mod == -1 ) {
3475: c = CM(m0);
3476: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3477: NEXTNM(mr0,mr);
3478: CM(mr) = _mulsf(CM(m),c);
3479: ndl_add(DL(m),d,DL(mr));
3480: }
3481: } else if ( mod ) {
1.16 noro 3482: c = CM(m0);
3483: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3484: NEXTNM(mr0,mr);
3485: c1 = CM(m);
1.19 noro 3486: DMAR(c1,c,0,mod,c2);
1.16 noro 3487: CM(mr) = c2;
3488: ndl_add(DL(m),d,DL(mr));
3489: }
3490: } else {
1.113 noro 3491: q = CQ(m0);
1.16 noro 3492: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3493: NEXTNM(mr0,mr);
1.113 noro 3494: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3495: ndl_add(DL(m),d,DL(mr));
3496: }
1.4 noro 3497: }
1.9 noro 3498: NEXT(mr) = 0;
1.31 noro 3499: MKND(NV(p),mr0,len,r);
1.34 noro 3500: SG(r) = SG(p) + TD(d);
1.9 noro 3501: return r;
1.4 noro 3502: }
3503: }
3504:
1.104 noro 3505: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3506: {
3507: NM mq0,mq;
1.102 noro 3508: NMV tm;
1.113 noro 3509: Q q;
1.104 noro 3510: int i,nv,sg,c,c1,c2,hindex;
3511: ND p,t,r;
3512: N tnm;
3513:
1.124 noro 3514: if ( bucket->m < 0 ) return 0;
1.99 noro 3515: else {
1.104 noro 3516: nv = NV(d);
1.99 noro 3517: mq0 = 0;
1.102 noro 3518: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3519: while ( 1 ) {
1.113 noro 3520: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3521: if ( hindex < 0 ) break;
3522: p = bucket->body[hindex];
1.99 noro 3523: NEXTNM(mq0,mq);
3524: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3525: ndl_copy(DL(mq),DL(tm));
3526: if ( mod ) {
3527: c1 = invm(HCM(d),mod); c2 = HCM(p);
3528: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3529: CM(tm) = mod-c;
3530: } else {
1.113 noro 3531: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3532: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3533: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3534: }
3535: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3536: bucket->body[hindex] = nd_remove_head(p);
3537: t = nd_remove_head(t);
3538: add_pbucket(mod,bucket,t);
3539: }
3540: if ( !mq0 )
3541: r = 0;
3542: else {
3543: NEXT(mq) = 0;
3544: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3545: MKND(nv,mq0,i,r);
3546: /* XXX */
3547: SG(r) = HTD(r);
1.99 noro 3548: }
3549: return r;
3550: }
3551: }
3552:
1.43 noro 3553: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3554: {
1.13 noro 3555: NMV m,mr,mr0,t;
3556: int len,i,k;
1.11 noro 3557:
1.61 noro 3558: if ( !p ) return;
3559: m = BDY(p); len = LEN(p);
3560: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3561: m = (NMV)((char *)mr0+(len-1)*oadv);
3562: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3563: t = (NMV)ALLOCA(nmv_adv);
3564: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3565: CQ(t) = CQ(m);
1.61 noro 3566: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3567: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3568: CQ(mr) = CQ(t);
1.61 noro 3569: ndl_copy(DL(t),DL(mr));
3570: }
3571: BDY(p) = mr0;
3572: }
3573:
3574: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3575: {
3576: NMV m,mr,mr0;
3577: int len,i;
3578: NDV r;
1.11 noro 3579:
1.61 noro 3580: if ( !p ) return 0;
3581: m = BDY(p); len = LEN(p);
3582: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3583: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3584: ndl_zero(DL(mr));
3585: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3586: CQ(mr) = CQ(m);
1.11 noro 3587: }
1.61 noro 3588: MKNDV(NV(p),mr0,len,r);
3589: SG(r) = SG(p);
3590: return r;
1.11 noro 3591: }
3592:
1.61 noro 3593: /* duplicate p */
3594:
3595: NDV ndv_dup(int mod,NDV p)
1.3 noro 3596: {
3597: NDV d;
1.61 noro 3598: NMV t,m,m0;
1.3 noro 3599: int i,len;
3600:
1.34 noro 3601: if ( !p ) return 0;
1.31 noro 3602: len = LEN(p);
1.34 noro 3603: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3604: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3605: ndl_copy(DL(t),DL(m));
1.113 noro 3606: CQ(m) = CQ(t);
1.3 noro 3607: }
3608: MKNDV(NV(p),m0,len,d);
1.23 noro 3609: SG(d) = SG(p);
3610: return d;
3611: }
3612:
1.63 noro 3613: ND nd_dup(ND p)
3614: {
3615: ND d;
3616: NM t,m,m0;
3617:
3618: if ( !p ) return 0;
3619: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3620: NEXTNM(m0,m);
3621: ndl_copy(DL(t),DL(m));
1.113 noro 3622: CQ(m) = CQ(t);
1.63 noro 3623: }
3624: if ( m0 ) NEXT(m) = 0;
3625: MKND(NV(p),m0,LEN(p),d);
3626: SG(d) = SG(p);
3627: return d;
3628: }
3629:
1.61 noro 3630: /* XXX if p->len == 0 then it represents 0 */
3631:
3632: void ndv_mod(int mod,NDV p)
3633: {
3634: NMV t,d;
1.125 noro 3635: int r,s,u;
1.61 noro 3636: int i,len,dlen;
1.71 noro 3637: Obj gfs;
1.61 noro 3638:
3639: if ( !p ) return;
3640: len = LEN(p);
3641: dlen = 0;
1.71 noro 3642: if ( mod == -1 )
3643: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3644: simp_ff((Obj)CP(t),&gfs);
3645: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3646: CM(d) = r;
3647: ndl_copy(DL(t),DL(d));
3648: NMV_ADV(d);
3649: dlen++;
3650: }
1.71 noro 3651: else
3652: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3653: r = rem(NM(CQ(t)),mod);
1.71 noro 3654: if ( r ) {
1.113 noro 3655: if ( SGN(CQ(t)) < 0 )
3656: r = mod-r;
1.125 noro 3657: if ( DN(CQ(t)) ) {
3658: s = rem(DN(CQ(t)),mod);
3659: if ( !s )
3660: error("ndv_mod : division by 0");
3661: s = invm(s,mod);
3662: DMAR(r,s,0,mod,u); r = u;
3663: }
1.71 noro 3664: CM(d) = r;
3665: ndl_copy(DL(t),DL(d));
3666: NMV_ADV(d);
3667: dlen++;
3668: }
3669: }
1.61 noro 3670: LEN(p) = dlen;
3671: }
3672:
3673: NDV ptondv(VL vl,VL dvl,P p)
3674: {
3675: ND nd;
3676:
3677: nd = ptond(vl,dvl,p);
3678: return ndtondv(0,nd);
3679: }
3680:
3681: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3682: {
1.61 noro 3683: int n,i,j,k,e;
3684: VL tvl;
3685: V v;
3686: DCP dc;
3687: DCP *w;
3688: ND r,s,t,u;
3689: P x;
3690: int c;
3691: UINT *d;
1.23 noro 3692: NM m,m0;
1.61 noro 3693:
3694: if ( !p )
3695: return 0;
3696: else if ( NUM(p) ) {
3697: NEWNM(m);
3698: ndl_zero(DL(m));
1.113 noro 3699: CQ(m) = (Q)p;
1.61 noro 3700: NEXT(m) = 0;
3701: MKND(nd_nvar,m,1,r);
3702: SG(r) = 0;
3703: return r;
3704: } else {
3705: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3706: w = (DCP *)ALLOCA(k*sizeof(DCP));
3707: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3708: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3709: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3710: if ( !tvl ) {
3711: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3712: t = ptond(vl,dvl,COEF(w[j]));
3713: pwrp(vl,x,DEG(w[j]),&p);
3714: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3715: }
3716: return s;
3717: } else {
3718: NEWNM(m0); d = DL(m0);
3719: for ( j = k-1, s = 0; j >= 0; j-- ) {
3720: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3721: TD(d) = MUL_WEIGHT(e,i);
3722: if ( nd_blockmask) ndl_weight_mask(d);
3723: t = ptond(vl,dvl,COEF(w[j]));
3724: for ( m = BDY(t); m; m = NEXT(m) )
3725: ndl_addto(DL(m),d);
3726: SG(t) += TD(d);
3727: s = nd_add(0,s,t);
3728: }
3729: FREENM(m0);
3730: return s;
3731: }
3732: }
3733: }
3734:
3735: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3736: {
3737: VL tvl;
3738: int len,n,j,i,e;
3739: NMV m;
3740: Q q;
3741: P c;
3742: UINT *d;
3743: P s,r,u,t,w;
1.71 noro 3744: GFS gfs;
1.23 noro 3745:
1.34 noro 3746: if ( !p ) return 0;
1.61 noro 3747: else {
3748: len = LEN(p);
3749: n = NV(p);
3750: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3751: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3752: if ( mod == -1 ) {
3753: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3754: } else if ( mod ) {
1.61 noro 3755: STOQ(CM(m),q); c = (P)q;
3756: } else
1.113 noro 3757: c = CP(m);
1.61 noro 3758: d = DL(m);
3759: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3760: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3761: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3762: }
3763: addp(vl,s,t,&u); s = u;
3764: }
3765: return s;
1.23 noro 3766: }
1.3 noro 3767: }
3768:
1.61 noro 3769: NDV ndtondv(int mod,ND p)
1.11 noro 3770: {
3771: NDV d;
1.61 noro 3772: NMV m,m0;
3773: NM t;
3774: int i,len;
1.11 noro 3775:
1.34 noro 3776: if ( !p ) return 0;
1.61 noro 3777: len = LEN(p);
1.74 noro 3778: if ( mod )
3779: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3780: else
3781: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3782: #if 0
1.74 noro 3783: ndv_alloc += nmv_adv*len;
1.103 noro 3784: #endif
1.61 noro 3785: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3786: ndl_copy(DL(t),DL(m));
1.113 noro 3787: CQ(m) = CQ(t);
1.11 noro 3788: }
1.61 noro 3789: MKNDV(NV(p),m0,len,d);
1.14 noro 3790: SG(d) = SG(p);
1.11 noro 3791: return d;
3792: }
3793:
1.61 noro 3794: ND ndvtond(int mod,NDV p)
1.11 noro 3795: {
1.61 noro 3796: ND d;
3797: NM m,m0;
1.11 noro 3798: NMV t;
1.61 noro 3799: int i,len;
1.11 noro 3800:
1.34 noro 3801: if ( !p ) return 0;
1.11 noro 3802: m0 = 0;
1.61 noro 3803: len = p->len;
3804: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3805: NEXTNM(m0,m);
3806: ndl_copy(DL(t),DL(m));
1.113 noro 3807: CQ(m) = CQ(t);
1.11 noro 3808: }
3809: NEXT(m) = 0;
1.61 noro 3810: MKND(NV(p),m0,len,d);
1.14 noro 3811: SG(d) = SG(p);
1.11 noro 3812: return d;
3813: }
3814:
1.3 noro 3815: void ndv_print(NDV p)
3816: {
3817: NMV m;
3818: int i,len;
3819:
1.34 noro 3820: if ( !p ) printf("0\n");
1.3 noro 3821: else {
1.14 noro 3822: len = LEN(p);
1.3 noro 3823: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3824: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3825: else printf("+%d*",CM(m));
1.16 noro 3826: ndl_print(DL(m));
3827: }
3828: printf("\n");
3829: }
3830: }
3831:
1.113 noro 3832: void ndv_print_q(NDV p)
1.16 noro 3833: {
3834: NMV m;
3835: int i,len;
3836:
1.34 noro 3837: if ( !p ) printf("0\n");
1.16 noro 3838: else {
3839: len = LEN(p);
3840: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3841: printf("+");
1.113 noro 3842: printexpr(CO,(Obj)CQ(m));
1.16 noro 3843: printf("*");
1.14 noro 3844: ndl_print(DL(m));
1.3 noro 3845: }
3846: printf("\n");
3847: }
1.25 noro 3848: }
3849:
1.61 noro 3850: NODE ndv_reducebase(NODE x)
1.27 noro 3851: {
3852: int len,i,j;
3853: NDV *w;
3854: NODE t,t0;
3855:
3856: len = length(x);
3857: w = (NDV *)ALLOCA(len*sizeof(NDV));
3858: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3859: for ( i = 0; i < len; i++ ) {
3860: for ( j = 0; j < i; j++ ) {
3861: if ( w[i] && w[j] )
3862: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3863: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3864: }
3865: }
3866: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3867: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3868: }
3869: NEXT(t) = 0; x = t0;
3870: return x;
1.11 noro 3871: }
1.32 noro 3872:
1.43 noro 3873: /* XXX incomplete */
3874:
1.32 noro 3875: void nd_init_ord(struct order_spec *ord)
3876: {
1.43 noro 3877: switch ( ord->id ) {
1.32 noro 3878: case 0:
1.43 noro 3879: switch ( ord->ord.simple ) {
3880: case 0:
3881: nd_dcomp = 1;
3882: nd_isrlex = 1;
3883: break;
3884: case 1:
3885: nd_dcomp = 1;
3886: nd_isrlex = 0;
3887: break;
3888: case 2:
3889: nd_dcomp = 0;
3890: nd_isrlex = 0;
1.45 noro 3891: ndl_compare_function = ndl_lex_compare;
1.58 noro 3892: break;
3893: case 11:
3894: /* XXX */
3895: nd_dcomp = 0;
3896: nd_isrlex = 1;
3897: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3898: break;
3899: default:
3900: error("nd_gr : unsupported order");
3901: }
1.32 noro 3902: break;
3903: case 1:
1.96 noro 3904: /* block order */
1.43 noro 3905: /* XXX */
3906: nd_dcomp = -1;
1.32 noro 3907: nd_isrlex = 0;
1.45 noro 3908: ndl_compare_function = ndl_block_compare;
1.34 noro 3909: break;
1.43 noro 3910: case 2:
1.96 noro 3911: /* matrix order */
3912: /* XXX */
3913: nd_dcomp = -1;
3914: nd_isrlex = 0;
3915: nd_matrix_len = ord->ord.matrix.row;
3916: nd_matrix = ord->ord.matrix.matrix;
3917: ndl_compare_function = ndl_matrix_compare;
3918: break;
3919: case 3:
1.97 noro 3920: /* composite order */
3921: nd_dcomp = -1;
3922: nd_isrlex = 0;
3923: nd_worb_len = ord->ord.composite.length;
3924: nd_worb = ord->ord.composite.w_or_b;
3925: ndl_compare_function = ndl_composite_compare;
1.32 noro 3926: break;
3927: }
1.41 noro 3928: nd_ord = ord;
1.32 noro 3929: }
3930:
1.43 noro 3931: BlockMask nd_create_blockmask(struct order_spec *ord)
3932: {
3933: int n,i,j,s,l;
1.61 noro 3934: UINT *t;
1.43 noro 3935: BlockMask bm;
3936:
1.96 noro 3937: /* we only create mask table for block order */
3938: if ( ord->id != 1 )
1.43 noro 3939: return 0;
3940: n = ord->ord.block.length;
3941: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3942: bm->n = n;
3943: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3944: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3945: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3946: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3947: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3948: l = bm->order_pair[i].length;
3949: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3950: }
3951: return bm;
1.57 noro 3952: }
3953:
3954: EPOS nd_create_epos(struct order_spec *ord)
3955: {
3956: int i,j,l,s,ord_l,ord_o;
3957: EPOS epos;
3958: struct order_pair *op;
3959:
3960: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3961: switch ( ord->id ) {
3962: case 0:
3963: if ( nd_isrlex ) {
3964: for ( i = 0; i < nd_nvar; i++ ) {
3965: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3966: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3967: }
3968: } else {
3969: for ( i = 0; i < nd_nvar; i++ ) {
3970: epos[i].i = nd_exporigin + i/nd_epw;
3971: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3972: }
3973: }
3974: break;
3975: case 1:
3976: /* block order */
3977: l = ord->ord.block.length;
3978: op = ord->ord.block.order_pair;
3979: for ( j = 0, s = 0; j < l; j++ ) {
3980: ord_o = op[j].order;
3981: ord_l = op[j].length;
3982: if ( !ord_o )
3983: for ( i = 0; i < ord_l; i++ ) {
3984: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3985: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3986: }
3987: else
3988: for ( i = 0; i < ord_l; i++ ) {
3989: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3990: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3991: }
3992: s += ord_l;
3993: }
3994: break;
3995: case 2:
1.96 noro 3996: /* matrix order */
3997: case 3:
3998: /* composite order */
3999: for ( i = 0; i < nd_nvar; i++ ) {
4000: epos[i].i = nd_exporigin + i/nd_epw;
4001: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4002: }
4003: break;
1.57 noro 4004: }
4005: return epos;
1.43 noro 4006: }
1.59 noro 4007:
4008: /* external interface */
4009:
4010: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4011: {
1.61 noro 4012: NODE t,in0,in;
1.59 noro 4013: ND nd,nf;
1.61 noro 4014: NDV ndv;
4015: VL vv,tv;
4016: int stat,nvar,max,e;
1.113 noro 4017: union oNDC dn;
1.59 noro 4018:
1.116 noro 4019: if ( !f ) {
4020: *rp = 0;
4021: return;
4022: }
1.59 noro 4023: pltovl(v,&vv);
1.74 noro 4024: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4025:
4026: /* get the degree bound */
4027: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4028: for ( tv = vv; tv; tv = NEXT(tv) ) {
4029: e = getdeg(tv->v,(P)BDY(t));
4030: max = MAX(e,max);
4031: }
4032: for ( tv = vv; tv; tv = NEXT(tv) ) {
4033: e = getdeg(tv->v,f);
4034: max = MAX(e,max);
4035: }
4036:
1.59 noro 4037: nd_init_ord(ord);
1.61 noro 4038: nd_setup_parameters(nvar,max);
4039:
4040: /* conversion to ndv */
4041: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4042: NEXTNODE(in0,in);
4043: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4044: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4045: }
4046: NEXTNODE(in0,in);
4047: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4048: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4049: NEXT(in) = 0;
4050:
1.128 noro 4051: /* dont sort */
4052: ndv_setup(m,0,in0,1);
1.59 noro 4053: nd_psn--;
4054: nd_scale=2;
4055: while ( 1 ) {
4056: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4057: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4058: if ( !stat ) {
4059: nd_psn++;
1.103 noro 4060: nd_reconstruct(0,0);
1.59 noro 4061: nd_psn--;
4062: } else
4063: break;
4064: }
1.61 noro 4065: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4066: }
4067:
4068: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4069: {
4070: NM m;
4071: UINT *t,*s;
4072: int i;
4073:
4074: for ( i = 0; i < n; i++ ) r[i] = 0;
4075: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4076: t = DL(m);
4077: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4078: r[i] = CM(m);
4079: }
4080: for ( i = 0; !r[i]; i++ );
4081: return i;
4082: }
4083:
1.113 noro 4084: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4085: {
1.107 noro 4086: NM m;
1.74 noro 4087: UINT *t,*s;
1.107 noro 4088: int i;
1.74 noro 4089:
4090: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4091: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4092: t = DL(m);
4093: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4094: r[i] = CQ(m);
1.74 noro 4095: }
4096: for ( i = 0; !r[i]; i++ );
4097: return i;
4098: }
4099:
1.129 noro 4100: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4101: {
4102: NM m;
4103: NMV mr;
4104: UINT *d,*t,*s;
4105: NDV p;
4106: int i,j,len;
4107: Q *r;
4108:
4109: m = pair->mul;
4110: d = DL(m);
4111: p = nd_ps[pair->index];
4112: len = LEN(p);
4113: r = (Q *)CALLOC(n,sizeof(Q));
4114: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4115: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4116: ndl_add(d,DL(mr),t);
4117: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4118: r[i] = CQ(mr);
4119: }
4120: return r;
4121: }
4122:
1.67 noro 4123: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4124: {
4125: NM m;
4126: NMV mr;
4127: UINT *d,*t,*s;
4128: NDV p;
1.67 noro 4129: unsigned char *ivc;
4130: unsigned short *ivs;
1.81 noro 4131: UINT *v,*ivi,*s0v;
1.67 noro 4132: int i,j,len,prev,diff,cdiff;
4133: IndArray r;
1.64 noro 4134:
4135: m = pair->mul;
4136: d = DL(m);
4137: p = nd_ps[pair->index];
4138: len = LEN(p);
4139: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4140: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4141: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4142: ndl_add(d,DL(mr),t);
4143: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4144: v[j] = i;
4145: }
1.81 noro 4146: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4147: r->head = v[0];
4148: diff = 0;
4149: for ( i = 1; i < len; i++ ) {
4150: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4151: }
4152: if ( diff < 256 ) {
4153: r->width = 1;
4154: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4155: r->index.c = ivc;
4156: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4157: } else if ( diff < 65536 ) {
4158: r->width = 2;
4159: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4160: r->index.s = ivs;
4161: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4162: } else {
4163: r->width = 4;
4164: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4165: r->index.i = ivi;
4166: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4167: }
1.67 noro 4168: return r;
1.64 noro 4169: }
4170:
1.133 noro 4171: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4172: {
4173: int i,j,k,len,pos,prev;
1.113 noro 4174: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4175: IndArray ivect;
4176: unsigned char *ivc;
4177: unsigned short *ivs;
4178: unsigned int *ivi;
4179: NDV redv;
4180: NMV mr;
4181: NODE rp;
4182: int maxrs;
1.133 noro 4183: double hmag;
4184: struct oVECT v;
1.107 noro 4185:
1.133 noro 4186: v.id = O_VECT; v.len = col; v.body = (pointer *)svect;
1.107 noro 4187: maxrs = 0;
1.134 ! noro 4188: for ( i = 0; i < col && !svect[i]; i++ );
! 4189: if ( i == col ) return maxrs;
! 4190: hmag = p_mag((P)svect[i])*nd_scale;
1.107 noro 4191: for ( i = 0; i < nred; i++ ) {
4192: ivect = imat[i];
4193: k = ivect->head;
4194: if ( svect[k] ) {
4195: maxrs = MAX(maxrs,rp0[i]->sugar);
1.133 noro 4196: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
1.107 noro 4197: len = LEN(redv); mr = BDY(redv);
1.113 noro 4198: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4199: chsgnq(cs,&mcs);
4200: if ( !UNIQ(cr) ) {
4201: for ( j = 0; j < col; j++ ) {
4202: mulq(svect[j],cr,&c1); svect[j] = c1;
4203: }
4204: }
1.107 noro 4205: svect[k] = 0; prev = k;
4206: switch ( ivect->width ) {
4207: case 1:
4208: ivc = ivect->index.c;
4209: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4210: pos = prev+ivc[j]; prev = pos;
1.113 noro 4211: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4212: }
4213: break;
4214: case 2:
4215: ivs = ivect->index.s;
4216: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4217: pos = prev+ivs[j]; prev = pos;
1.113 noro 4218: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4219: }
4220: break;
4221: case 4:
4222: ivi = ivect->index.i;
4223: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4224: pos = prev+ivi[j]; prev = pos;
1.113 noro 4225: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4226: }
4227: break;
4228: }
1.134 ! noro 4229: for ( j = k+1; j < col && !svect[j]; j++ );
! 4230: if ( j == col ) break;
! 4231: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
! 4232: igcdv(&v,&t);
! 4233: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
! 4234: }
1.107 noro 4235: }
1.133 noro 4236: }
4237: igcdv(&v,&t);
4238: if ( DP_Print ) {
4239: fprintf(asir_out,"-"); fflush(asir_out);
1.107 noro 4240: }
4241: return maxrs;
4242: }
4243:
1.76 noro 4244: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4245: {
1.67 noro 4246: int i,j,k,len,pos,prev;
1.66 noro 4247: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4248: IndArray ivect;
4249: unsigned char *ivc;
4250: unsigned short *ivs;
4251: unsigned int *ivi;
1.65 noro 4252: NDV redv;
1.67 noro 4253: NMV mr;
1.65 noro 4254: NODE rp;
1.76 noro 4255: int maxrs;
1.65 noro 4256:
1.76 noro 4257: maxrs = 0;
1.74 noro 4258: for ( i = 0; i < nred; i++ ) {
1.65 noro 4259: ivect = imat[i];
1.67 noro 4260: k = ivect->head; svect[k] %= m;
1.65 noro 4261: if ( c = svect[k] ) {
1.76 noro 4262: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4263: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4264: len = LEN(redv); mr = BDY(redv);
4265: svect[k] = 0; prev = k;
4266: switch ( ivect->width ) {
4267: case 1:
4268: ivc = ivect->index.c;
4269: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4270: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4271: prev = pos;
4272: DMA(c1,c,c2,up,lo);
4273: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4274: } else svect[pos] = lo;
4275: }
4276: break;
4277: case 2:
4278: ivs = ivect->index.s;
4279: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4280: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4281: prev = pos;
4282: DMA(c1,c,c2,up,lo);
4283: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4284: } else svect[pos] = lo;
4285: }
4286: break;
4287: case 4:
4288: ivi = ivect->index.i;
4289: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4290: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4291: prev = pos;
4292: DMA(c1,c,c2,up,lo);
4293: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4294: } else svect[pos] = lo;
4295: }
4296: break;
1.65 noro 4297: }
4298: }
4299: }
1.66 noro 4300: for ( i = 0; i < col; i++ )
4301: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4302: return maxrs;
1.65 noro 4303: }
4304:
1.76 noro 4305: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4306: {
4307: int i,j,k,len,pos,prev;
4308: UINT c,c1,c2,c3,up,lo,dmy;
4309: IndArray ivect;
4310: unsigned char *ivc;
4311: unsigned short *ivs;
4312: unsigned int *ivi;
4313: NDV redv;
4314: NMV mr;
4315: NODE rp;
1.76 noro 4316: int maxrs;
1.72 noro 4317:
1.76 noro 4318: maxrs = 0;
1.74 noro 4319: for ( i = 0; i < nred; i++ ) {
1.72 noro 4320: ivect = imat[i];
4321: k = ivect->head; svect[k] %= m;
4322: if ( c = svect[k] ) {
1.76 noro 4323: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4324: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4325: len = LEN(redv); mr = BDY(redv);
4326: svect[k] = 0; prev = k;
4327: switch ( ivect->width ) {
4328: case 1:
4329: ivc = ivect->index.c;
4330: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4331: pos = prev+ivc[j]; prev = pos;
4332: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4333: }
4334: break;
4335: case 2:
4336: ivs = ivect->index.s;
4337: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4338: pos = prev+ivs[j]; prev = pos;
4339: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4340: }
4341: break;
4342: case 4:
4343: ivi = ivect->index.i;
4344: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4345: pos = prev+ivi[j]; prev = pos;
4346: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4347: }
4348: break;
4349: }
4350: }
4351: }
1.76 noro 4352: return maxrs;
1.72 noro 4353: }
4354:
1.65 noro 4355: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4356: {
4357: int j,k,len;
4358: UINT *p;
4359: UINT c;
4360: NDV r;
4361: NMV mr0,mr;
4362:
4363: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4364: if ( !len ) return 0;
4365: else {
1.74 noro 4366: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4367: #if 0
1.74 noro 4368: ndv_alloc += nmv_adv*len;
1.103 noro 4369: #endif
1.65 noro 4370: mr = mr0;
4371: p = s0vect;
4372: for ( j = k = 0; j < col; j++, p += nd_wpd )
4373: if ( !rhead[j] ) {
4374: if ( c = vect[k++] ) {
4375: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4376: }
4377: }
4378: MKNDV(nd_nvar,mr0,len,r);
4379: return r;
4380: }
4381: }
4382:
1.129 noro 4383: /* for preprocessed vector */
4384:
1.113 noro 4385: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4386: {
4387: int j,k,len;
4388: UINT *p;
1.113 noro 4389: Q c;
1.107 noro 4390: NDV r;
4391: NMV mr0,mr;
4392:
4393: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4394: if ( !len ) return 0;
4395: else {
1.109 noro 4396: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4397: #if 0
4398: ndv_alloc += nmv_adv*len;
4399: #endif
4400: mr = mr0;
4401: p = s0vect;
4402: for ( j = k = 0; j < col; j++, p += nd_wpd )
4403: if ( !rhead[j] ) {
4404: if ( c = vect[k++] ) {
1.113 noro 4405: if ( DN(c) )
4406: error("afo");
4407: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4408: }
4409: }
4410: MKNDV(nd_nvar,mr0,len,r);
4411: return r;
4412: }
4413: }
4414:
1.129 noro 4415: /* for plain vector */
4416:
4417: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4418: {
4419: int j,k,len;
4420: UINT *p;
4421: Q c;
4422: NDV r;
4423: NMV mr0,mr;
4424:
4425: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4426: if ( !len ) return 0;
4427: else {
4428: mr0 = (NMV)GC_malloc(nmv_adv*len);
4429: #if 0
4430: ndv_alloc += nmv_adv*len;
4431: #endif
4432: mr = mr0;
4433: p = s0vect;
4434: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4435: if ( c = vect[k] ) {
4436: if ( DN(c) )
4437: error("afo");
4438: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4439: }
4440: MKNDV(nd_nvar,mr0,len,r);
4441: return r;
4442: }
4443: }
4444:
1.133 noro 4445: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 4446: {
4447: ND_pairs t;
4448: NODE sp0,sp;
4449: int stat;
4450: ND spol;
4451:
4452: for ( t = l; t; t = NEXT(t) ) {
1.133 noro 4453: stat = nd_sp(m,trace,t,&spol);
1.65 noro 4454: if ( !stat ) return 0;
4455: if ( spol ) {
4456: add_pbucket_symbolic(bucket,spol);
4457: }
4458: }
1.68 noro 4459: return 1;
1.65 noro 4460: }
4461:
1.133 noro 4462: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 4463: {
4464: NODE rp0,rp;
4465: NM mul,head,s0,s;
1.76 noro 4466: int index,col,i,sugar;
1.65 noro 4467: RHist h;
4468: UINT *s0v,*p;
4469: NM_ind_pair pair;
4470: ND red;
1.133 noro 4471: NDV *ps;
1.65 noro 4472:
4473: s0 = 0; rp0 = 0; col = 0;
1.133 noro 4474: ps = trace?nd_ps_trace:nd_ps;
1.65 noro 4475: while ( 1 ) {
4476: head = remove_head_pbucket_symbolic(bucket);
4477: if ( !head ) break;
4478: if ( !s0 ) s0 = head;
4479: else NEXT(s) = head;
4480: s = head;
4481: index = ndl_find_reducer(DL(head));
4482: if ( index >= 0 ) {
4483: h = nd_psh[index];
4484: NEWNM(mul);
4485: ndl_sub(DL(head),DL(h),DL(mul));
4486: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.133 noro 4487: sugar = TD(DL(mul))+SG(ps[index]);
1.76 noro 4488: MKNM_ind_pair(pair,mul,index,sugar);
1.133 noro 4489: red = ndv_mul_nm_symbolic(mul,ps[index]);
1.65 noro 4490: add_pbucket_symbolic(bucket,nd_remove_head(red));
4491: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4492: }
4493: col++;
4494: }
1.72 noro 4495: if ( rp0 ) NEXT(rp) = 0;
4496: NEXT(s) = 0;
1.65 noro 4497: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4498: for ( i = 0, p = s0v, s = s0; i < col;
4499: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4500: *s0vect = s0v;
4501: *r = rp0;
4502: return col;
4503: }
4504:
1.69 noro 4505: NODE nd_f4(int m)
4506: {
4507: int i,nh,stat,index;
4508: NODE r,g;
4509: ND_pairs d,l,t;
4510: ND spol,red;
4511: NDV nf,redv;
4512: NM s0,s;
1.84 noro 4513: NODE rp0,srp0,nflist;
1.69 noro 4514: int nsp,nred,col,rank,len,k,j,a;
4515: UINT c;
1.74 noro 4516: UINT **spmat;
1.69 noro 4517: UINT *s0vect,*svect,*p,*v;
4518: int *colstat;
4519: IndArray *imat;
4520: int *rhead;
4521: int spcol,sprow;
4522: int sugar;
4523: PGeoBucket bucket;
4524: struct oEGT eg0,eg1,eg_f4;
4525:
1.103 noro 4526: #if 0
1.74 noro 4527: ndv_alloc = 0;
1.103 noro 4528: #endif
1.69 noro 4529: g = 0; d = 0;
4530: for ( i = 0; i < nd_psn; i++ ) {
4531: d = update_pairs(d,g,i);
4532: g = update_base(g,i);
4533: }
4534: while ( d ) {
4535: get_eg(&eg0);
4536: l = nd_minsugarp(d,&d);
4537: sugar = SG(l);
4538: bucket = create_pbucket();
1.133 noro 4539: stat = nd_sp_f4(m,0,l,bucket);
1.69 noro 4540: if ( !stat ) {
4541: for ( t = l; NEXT(t); t = NEXT(t) );
4542: NEXT(t) = d; d = l;
1.103 noro 4543: d = nd_reconstruct(0,d);
1.69 noro 4544: continue;
4545: }
1.86 noro 4546: if ( bucket->m < 0 ) continue;
1.133 noro 4547: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
1.69 noro 4548: if ( !col ) {
4549: for ( t = l; NEXT(t); t = NEXT(t) );
4550: NEXT(t) = d; d = l;
1.103 noro 4551: d = nd_reconstruct(0,d);
1.69 noro 4552: continue;
4553: }
4554: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4555: if ( DP_Print )
4556: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4557: sugar,eg_f4.exectime+eg_f4.gctime);
4558: if ( 1 )
1.133 noro 4559: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
1.74 noro 4560: else
1.133 noro 4561: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
1.69 noro 4562: /* adding new bases */
1.74 noro 4563: for ( r = nflist; r; r = NEXT(r) ) {
4564: nf = (NDV)BDY(r);
1.69 noro 4565: ndv_removecont(m,nf);
1.129 noro 4566: if ( !m && nd_nalg ) {
4567: ND nf1;
4568:
4569: nf1 = ndvtond(m,nf);
4570: nd_monic(0,&nf1);
4571: nd_removecont(m,nf1);
4572: nf = ndtondv(m,nf1);
4573: }
1.77 noro 4574: nh = ndv_newps(m,nf,0);
1.69 noro 4575: d = update_pairs(d,g,nh);
4576: g = update_base(g,nh);
4577: }
4578: }
4579: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4580: #if 0
1.74 noro 4581: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4582: #endif
1.69 noro 4583: return g;
4584: }
1.74 noro 4585:
1.133 noro 4586: NODE nd_f4_trace(int m)
4587: {
4588: int i,nh,stat,index;
4589: NODE r,g;
4590: ND_pairs d,l,l0,t;
4591: ND spol,red;
4592: NDV nf,redv,nfqv,nfv;
4593: NM s0,s;
4594: NODE rp0,srp0,nflist;
4595: int nsp,nred,col,rank,len,k,j,a;
4596: UINT c;
4597: UINT **spmat;
4598: UINT *s0vect,*svect,*p,*v;
4599: int *colstat;
4600: IndArray *imat;
4601: int *rhead;
4602: int spcol,sprow;
4603: int sugar;
4604: PGeoBucket bucket;
4605: struct oEGT eg0,eg1,eg_f4;
4606:
4607: g = 0; d = 0;
4608: for ( i = 0; i < nd_psn; i++ ) {
4609: d = update_pairs(d,g,i);
4610: g = update_base(g,i);
4611: }
4612: while ( d ) {
4613: get_eg(&eg0);
4614: l = nd_minsugarp(d,&d);
4615: sugar = SG(l);
4616: bucket = create_pbucket();
4617: stat = nd_sp_f4(m,0,l,bucket);
4618: if ( !stat ) {
4619: for ( t = l; NEXT(t); t = NEXT(t) );
4620: NEXT(t) = d; d = l;
4621: d = nd_reconstruct(1,d);
4622: continue;
4623: }
4624: if ( bucket->m < 0 ) continue;
4625: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
4626: if ( !col ) {
4627: for ( t = l; NEXT(t); t = NEXT(t) );
4628: NEXT(t) = d; d = l;
4629: d = nd_reconstruct(1,d);
4630: continue;
4631: }
4632: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4633: if ( DP_Print )
4634: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4635: sugar,eg_f4.exectime+eg_f4.gctime);
4636: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
4637: if ( !l0 ) continue;
4638: l = l0;
4639:
4640: /* over Q */
4641: bucket = create_pbucket();
4642: stat = nd_sp_f4(0,1,l,bucket);
4643: if ( !stat ) {
4644: for ( t = l; NEXT(t); t = NEXT(t) );
4645: NEXT(t) = d; d = l;
4646: d = nd_reconstruct(1,d);
4647: continue;
4648: }
4649: if ( bucket->m < 0 ) continue;
4650: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
4651: if ( !col ) {
4652: for ( t = l; NEXT(t); t = NEXT(t) );
4653: NEXT(t) = d; d = l;
4654: d = nd_reconstruct(1,d);
4655: continue;
4656: }
4657: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
4658: /* adding new bases */
4659: for ( r = nflist; r; r = NEXT(r) ) {
4660: nfqv = (NDV)BDY(r);
4661: ndv_removecont(0,nfqv);
4662: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
4663: if ( nd_nalg ) {
4664: ND nf1;
4665:
4666: nf1 = ndvtond(m,nfqv);
4667: nd_monic(0,&nf1);
4668: nd_removecont(0,nf1);
4669: nfqv = ndtondv(0,nf1); nd_free(nf1);
4670: }
4671: nfv = ndv_dup(0,nfqv);
4672: ndv_mod(m,nfv);
4673: ndv_removecont(m,nfv);
4674: nh = ndv_newps(0,nfv,nfqv);
4675: d = update_pairs(d,g,nh);
4676: g = update_base(g,nh);
4677: }
4678: }
4679: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
4680: #if 0
4681: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
4682: #endif
4683: return g;
4684: }
4685:
4686: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 4687: {
1.67 noro 4688: IndArray *imat;
1.106 noro 4689: int nsp,nred,i;
1.65 noro 4690: int *rhead;
1.106 noro 4691: NODE r0,rp;
1.74 noro 4692: ND_pairs sp;
4693: NM_ind_pair *rvect;
1.63 noro 4694:
1.74 noro 4695: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4696: nred = length(rp0);
1.74 noro 4697: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4698: rhead = (int *)ALLOCA(col*sizeof(int));
4699: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4700:
1.74 noro 4701: /* construction of index arrays */
4702: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4703: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4704: rvect[i] = (NM_ind_pair)BDY(rp);
4705: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4706: rhead[imat[i]->head] = 1;
4707: }
1.107 noro 4708: if ( m )
1.133 noro 4709: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
1.107 noro 4710: else
1.133 noro 4711: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4712: return r0;
4713: }
1.74 noro 4714:
1.106 noro 4715: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 4716: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 4717: {
4718: int spcol,sprow,a;
4719: int i,j,k,l,rank;
4720: NODE r0,r;
4721: ND_pairs sp;
4722: ND spol;
4723: int **spmat;
4724: UINT *svect,*v;
4725: int *colstat;
4726: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4727: int maxrs;
4728: int *spsugar;
1.133 noro 4729: ND_pairs *spactive;
1.106 noro 4730:
4731: spcol = col-nred;
4732: get_eg(&eg0);
1.74 noro 4733: /* elimination (1st step) */
4734: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4735: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4736: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.133 noro 4737: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
1.74 noro 4738: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4739: nd_sp(m,0,sp,&spol);
1.75 noro 4740: if ( !spol ) continue;
1.74 noro 4741: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4742: if ( m == -1 )
4743: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4744: else
4745: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4746: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4747: if ( i < col ) {
4748: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4749: for ( j = k = 0; j < col; j++ )
4750: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4751: spsugar[sprow] = MAX(maxrs,SG(spol));
1.133 noro 4752: if ( nz )
4753: spactive[sprow] = sp;
1.74 noro 4754: sprow++;
4755: }
1.76 noro 4756: nd_free(spol);
1.74 noro 4757: }
1.80 noro 4758: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4759: if ( DP_Print ) {
4760: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4761: fflush(asir_out);
4762: }
1.74 noro 4763: /* free index arrays */
4764: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4765:
4766: /* elimination (2nd step) */
4767: colstat = (int *)ALLOCA(spcol*sizeof(int));
4768: if ( m == -1 )
1.76 noro 4769: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4770: else
1.133 noro 4771: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
1.74 noro 4772: r0 = 0;
4773: for ( i = 0; i < rank; i++ ) {
4774: NEXTNODE(r0,r); BDY(r) =
4775: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4776: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4777: GC_free(spmat[i]);
4778: }
1.109 noro 4779: if ( r0 ) NEXT(r) = 0;
1.133 noro 4780:
1.74 noro 4781: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4782: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4783: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4784: if ( DP_Print ) {
1.80 noro 4785: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4786: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4787: nsp,nred,sprow,spcol,rank);
4788: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4789: }
1.133 noro 4790: if ( nz ) {
4791: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
4792: if ( rank > 0 ) {
4793: NEXT(spactive[rank-1]) = 0;
4794: *nz = spactive[0];
4795: } else
4796: *nz = 0;
4797: }
1.74 noro 4798: return r0;
4799: }
4800:
1.133 noro 4801: #if 1
4802: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 4803: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4804: {
4805: int spcol,sprow,a;
4806: int i,j,k,l,rank;
4807: NODE r0,r;
4808: ND_pairs sp;
4809: ND spol;
1.113 noro 4810: Q **spmat;
4811: Q *svect,*v;
1.107 noro 4812: int *colstat;
4813: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4814: int maxrs;
4815: int *spsugar;
4816:
4817: spcol = col-nred;
4818: get_eg(&eg0);
4819: /* elimination (1st step) */
1.113 noro 4820: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4821: svect = (Q *)ALLOCA(col*sizeof(Q));
4822: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4823: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
1.133 noro 4824: nd_sp(0,trace,sp,&spol);
1.107 noro 4825: if ( !spol ) continue;
1.113 noro 4826: nd_to_vect_q(s0vect,col,spol,svect);
1.133 noro 4827: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
1.107 noro 4828: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4829: if ( i < col ) {
1.113 noro 4830: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4831: for ( j = k = 0; j < col; j++ )
4832: if ( !rhead[j] ) v[k++] = svect[j];
4833: spsugar[sprow] = MAX(maxrs,SG(spol));
4834: sprow++;
4835: }
1.109 noro 4836: /* nd_free(spol); */
1.107 noro 4837: }
4838: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4839: if ( DP_Print ) {
4840: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4841: fflush(asir_out);
4842: }
4843: /* free index arrays */
1.109 noro 4844: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4845:
4846: /* elimination (2nd step) */
4847: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4848: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.107 noro 4849: r0 = 0;
4850: for ( i = 0; i < rank; i++ ) {
4851: NEXTNODE(r0,r); BDY(r) =
1.113 noro 4852: (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
1.107 noro 4853: SG((NDV)BDY(r)) = spsugar[i];
1.109 noro 4854: /* GC_free(spmat[i]); */
1.107 noro 4855: }
1.109 noro 4856: if ( r0 ) NEXT(r) = 0;
4857:
4858: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4859: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4860: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4861: if ( DP_Print ) {
4862: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4863: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4864: nsp,nred,sprow,spcol,rank);
4865: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4866: }
4867: return r0;
4868: }
1.129 noro 4869: #else
4870: void printm(Q **mat,int row,int col)
4871: {
4872: int i,j;
4873: printf("[");
4874: for ( i = 0; i < row; i++ ) {
4875: for ( j = 0; j < col; j++ ) {
4876: printexpr(CO,mat[i][j]); printf(" ");
4877: }
4878: printf("]\n");
4879: }
4880: }
4881:
4882: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
4883: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4884: {
4885: int row,a;
4886: int i,j,rank;
4887: NODE r0,r;
4888: ND_pairs sp;
4889: ND spol;
4890: Q **mat;
4891: int *colstat;
4892: int *sugar;
4893:
4894: row = nsp+nred;
4895: /* make the matrix */
4896: mat = (Q **)ALLOCA(row*sizeof(Q *));
4897: sugar = (int *)ALLOCA(row*sizeof(int));
4898: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4899: nd_sp(0,0,sp,&spol);
4900: if ( !spol ) continue;
4901: mat[row] = (Q *)MALLOC(col*sizeof(Q));
4902: nd_to_vect_q(s0vect,col,spol,mat[row]);
4903: sugar[row] = SG(spol);
4904: row++;
4905: }
4906: for ( i = 0; i < nred; i++, row++ ) {
4907: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
4908: sugar[row] = rvect[i]->sugar;
4909: }
4910: /* elimination */
4911: colstat = (int *)ALLOCA(col*sizeof(int));
4912: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
4913: r0 = 0;
4914: for ( i = 0; i < rank; i++ ) {
4915: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
4916: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
4917: if ( rhead[j] ) continue;
4918: NEXTNODE(r0,r); BDY(r) =
4919: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
4920: SG((NDV)BDY(r)) = sugar[i];
4921: }
4922: if ( r0 ) NEXT(r) = 0;
4923: printf("\n");
4924: return r0;
4925: }
4926: #endif
1.107 noro 4927:
1.74 noro 4928: FILE *nd_write,*nd_read;
4929:
4930: void nd_send_int(int a) {
4931: write_int(nd_write,&a);
4932: }
4933:
4934: void nd_send_intarray(int *p,int len) {
4935: write_intarray(nd_write,p,len);
4936: }
4937:
4938: int nd_recv_int() {
4939: int a;
4940:
4941: read_int(nd_read,&a);
4942: return a;
4943: }
4944:
4945: void nd_recv_intarray(int *p,int len) {
4946: read_intarray(nd_read,p,len);
4947: }
4948:
4949: void nd_send_ndv(NDV p) {
4950: int len,i;
4951: NMV m;
4952:
4953: if ( !p ) nd_send_int(0);
4954: else {
4955: len = LEN(p);
4956: nd_send_int(len);
4957: m = BDY(p);
4958: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4959: nd_send_int(CM(m));
4960: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4961: }
1.74 noro 4962: }
4963: }
4964:
4965: void nd_send_nd(ND p) {
4966: int len,i;
4967: NM m;
4968:
4969: if ( !p ) nd_send_int(0);
4970: else {
4971: len = LEN(p);
4972: nd_send_int(len);
4973: m = BDY(p);
4974: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4975: nd_send_int(CM(m));
4976: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4977: }
1.74 noro 4978: }
4979: }
1.65 noro 4980:
1.74 noro 4981: NDV nd_recv_ndv()
4982: {
4983: int len,i;
4984: NMV m,m0;
4985: NDV r;
1.65 noro 4986:
1.74 noro 4987: len = nd_recv_int();
4988: if ( !len ) return 0;
4989: else {
4990: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4991: #if 0
1.74 noro 4992: ndv_alloc += len*nmv_adv;
1.103 noro 4993: #endif
1.74 noro 4994: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4995: CM(m) = nd_recv_int();
4996: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4997: }
1.74 noro 4998: MKNDV(nd_nvar,m0,len,r);
4999: return r;
5000: }
5001: }
1.65 noro 5002:
1.74 noro 5003: int ox_exec_f4_red(Q proc)
5004: {
5005: Obj obj;
5006: STRING fname;
5007: NODE arg;
5008: int s;
5009: extern int ox_need_conv,ox_file_io;
5010:
5011: MKSTR(fname,"nd_exec_f4_red");
5012: arg = mknode(2,proc,fname);
5013: Pox_cmo_rpc(arg,&obj);
5014: s = get_ox_server_id(QTOS(proc));
5015: nd_write = iofp[s].out;
5016: nd_read = iofp[s].in;
5017: ox_need_conv = ox_file_io = 0;
5018: return s;
5019: }
5020:
1.133 noro 5021: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5022: {
5023: int nsp,nred;
5024: int i,rank,s;
5025: NODE rp,r0,r;
5026: ND_pairs sp;
5027: NM_ind_pair pair;
5028: NMV nmv;
5029: NM nm;
5030: NDV nf;
5031: Obj proc,dmy;
5032:
5033: ox_launch_main(0,0,&proc);
5034: s = ox_exec_f4_red((Q)proc);
5035:
5036: nd_send_int(m);
5037: nd_send_int(nd_nvar);
5038: nd_send_int(nd_bpe);
5039: nd_send_int(nd_wpd);
5040: nd_send_int(nmv_adv);
5041:
1.89 noro 5042: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 5043:
5044: nd_send_int(nd_psn);
5045: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5046:
5047: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5048: nd_send_int(nsp);
5049: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5050: nd_send_int(sp->i1); nd_send_int(sp->i2);
5051: }
5052:
5053: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5054:
5055: nred = length(rp0); nd_send_int(nred);
5056: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5057: pair = (NM_ind_pair)BDY(rp);
5058: nd_send_int(pair->index);
5059: nd_send_intarray(pair->mul->dl,nd_wpd);
5060: }
5061: fflush(nd_write);
5062: rank = nd_recv_int();
5063: fprintf(asir_out,"rank=%d\n",rank);
5064: r0 = 0;
5065: for ( i = 0; i < rank; i++ ) {
5066: nf = nd_recv_ndv();
5067: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5068: }
5069: Pox_shutdown(mknode(1,proc),&dmy);
5070: return r0;
5071: }
5072:
5073: /* server side */
5074:
5075: void nd_exec_f4_red_dist()
5076: {
5077: int m,i,nsp,col,s0size,nred,spcol,j,k;
5078: NM_ind_pair *rp0;
5079: NDV nf;
5080: UINT *s0vect;
5081: IndArray *imat;
5082: int *rhead;
5083: int **spmat;
5084: UINT *svect,*v;
5085: ND_pairs *sp0;
5086: int *colstat;
5087: int a,sprow,rank;
1.89 noro 5088: struct order_spec *ord;
1.74 noro 5089: Obj ordspec;
5090: ND spol;
1.76 noro 5091: int maxrs;
5092: int *spsugar;
1.74 noro 5093:
5094: nd_read = iofp[0].in;
5095: nd_write = iofp[0].out;
5096: m = nd_recv_int();
5097: nd_nvar = nd_recv_int();
5098: nd_bpe = nd_recv_int();
5099: nd_wpd = nd_recv_int();
5100: nmv_adv = nd_recv_int();
5101:
5102: loadobj(nd_read,&ordspec);
1.89 noro 5103: create_order_spec(0,ordspec,&ord);
5104: nd_init_ord(ord);
1.74 noro 5105: nd_setup_parameters(nd_nvar,0);
5106:
5107: nd_psn = nd_recv_int();
5108: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5109: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5110: for ( i = 0; i < nd_psn; i++ ) {
5111: nd_ps[i] = nd_recv_ndv();
5112: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5113: }
5114:
5115: nsp = nd_recv_int();
5116: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5117: for ( i = 0; i < nsp; i++ ) {
5118: NEWND_pairs(sp0[i]);
5119: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5120: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5121: }
5122:
5123: col = nd_recv_int();
5124: s0size = col*nd_wpd;
5125: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5126: nd_recv_intarray(s0vect,s0size);
5127:
5128: nred = nd_recv_int();
5129: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5130: for ( i = 0; i < nred; i++ ) {
5131: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5132: rp0[i]->index = nd_recv_int();
5133: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5134: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5135: }
5136:
5137: spcol = col-nred;
5138: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5139: rhead = (int *)MALLOC(col*sizeof(int));
5140: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5141:
5142: /* construction of index arrays */
5143: for ( i = 0; i < nred; i++ ) {
5144: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5145: rhead[imat[i]->head] = 1;
5146: }
5147:
5148: /* elimination (1st step) */
5149: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5150: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5151: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5152: for ( a = sprow = 0; a < nsp; a++ ) {
5153: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5154: if ( !spol ) continue;
1.74 noro 5155: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5156: if ( m == -1 )
5157: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5158: else
5159: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5160: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5161: if ( i < col ) {
5162: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5163: for ( j = k = 0; j < col; j++ )
5164: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5165: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5166: sprow++;
5167: }
1.76 noro 5168: nd_free(spol);
1.74 noro 5169: }
5170: /* elimination (2nd step) */
5171: colstat = (int *)ALLOCA(spcol*sizeof(int));
5172: if ( m == -1 )
1.76 noro 5173: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5174: else
1.133 noro 5175: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
1.74 noro 5176: nd_send_int(rank);
5177: for ( i = 0; i < rank; i++ ) {
5178: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5179: nd_send_ndv(nf);
1.63 noro 5180: }
1.74 noro 5181: fflush(nd_write);
1.107 noro 5182: }
5183:
1.113 noro 5184: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5185: {
1.109 noro 5186: int mod,i,j,t,c,rank,rank0,inv;
5187: int *ci,*ri;
1.113 noro 5188: Q dn;
1.109 noro 5189: MAT m,nm;
1.108 noro 5190: int **wmat;
5191:
5192: /* XXX */
5193: mod = 99999989;
5194: wmat = (int **)ALLOCA(row*sizeof(int *));
5195: for ( i = 0; i < row; i++ ) {
5196: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5197: for ( j = 0; j < col; j++ ) {
1.113 noro 5198: if ( mat0[i][j] ) {
5199: t = rem(NM(mat0[i][j]),mod);
5200: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5201: wmat[i][j] = t;
5202: } else
1.108 noro 5203: wmat[i][j] = 0;
5204: }
5205: }
1.133 noro 5206: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
1.109 noro 5207: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5208: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5209: if ( rank != rank0 )
5210: error("afo");
5211: for ( i = 0; i < row; i++ )
5212: for ( j = 0; j < col; j++ )
5213: mat0[i][j] = 0;
5214: c = col-rank;
5215: for ( i = 0; i < rank; i++ ) {
1.113 noro 5216: mat0[i][ri[i]] = dn;
1.109 noro 5217: for ( j = 0; j < c; j++ )
1.113 noro 5218: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5219: }
1.113 noro 5220: inv = invm(rem(NM(dn),mod),mod);
5221: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5222: for ( i = 0; i < row; i++ )
5223: for ( j = 0; j < col; j++ ) {
1.113 noro 5224: if ( mat0[i][j] ) {
5225: t = rem(NM(mat0[i][j]),mod);
5226: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5227: } else
1.109 noro 5228: t = 0;
5229: c = dmar(t,inv,0,mod);
5230: if ( wmat[i][j] != c )
5231: error("afo");
5232: }
5233: return rank;
1.76 noro 5234: }
5235:
1.133 noro 5236: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5237: {
5238: int i,j,k,l,inv,a,rank,s;
5239: unsigned int *t,*pivot,*pk;
5240: unsigned int **mat;
1.133 noro 5241: ND_pairs pair;
1.76 noro 5242:
5243: mat = (unsigned int **)mat0;
5244: for ( rank = 0, j = 0; j < col; j++ ) {
5245: for ( i = rank; i < row; i++ )
5246: mat[i][j] %= md;
5247: for ( i = rank; i < row; i++ )
5248: if ( mat[i][j] )
5249: break;
5250: if ( i == row ) {
5251: colstat[j] = 0;
5252: continue;
5253: } else
5254: colstat[j] = 1;
5255: if ( i != rank ) {
5256: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5257: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
1.133 noro 5258: if ( spactive ) {
5259: pair = spactive[i]; spactive[i] = spactive[rank];
5260: spactive[rank] = pair;
5261: }
1.76 noro 5262: }
5263: pivot = mat[rank];
5264: s = sugar[rank];
5265: inv = invm(pivot[j],md);
5266: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5267: if ( *pk ) {
5268: if ( *pk >= (unsigned int)md )
5269: *pk %= md;
5270: DMAR(*pk,inv,0,md,*pk)
5271: }
5272: for ( i = rank+1; i < row; i++ ) {
5273: t = mat[i];
5274: if ( a = t[j] ) {
5275: sugar[i] = MAX(sugar[i],s);
5276: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5277: }
5278: }
5279: rank++;
5280: }
5281: for ( j = col-1, l = rank-1; j >= 0; j-- )
5282: if ( colstat[j] ) {
5283: pivot = mat[l];
5284: s = sugar[l];
5285: for ( i = 0; i < l; i++ ) {
5286: t = mat[i];
5287: t[j] %= md;
5288: if ( a = t[j] ) {
5289: sugar[i] = MAX(sugar[i],s);
5290: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5291: }
5292: }
5293: l--;
5294: }
5295: for ( j = 0, l = 0; l < rank; j++ )
5296: if ( colstat[j] ) {
5297: t = mat[l];
5298: for ( k = j; k < col; k++ )
5299: if ( t[k] >= (unsigned int)md )
5300: t[k] %= md;
5301: l++;
5302: }
5303: return rank;
5304: }
5305:
5306: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5307: {
5308: int i,j,k,l,inv,a,rank,s;
5309: unsigned int *t,*pivot,*pk;
5310: unsigned int **mat;
5311:
5312: mat = (unsigned int **)mat0;
5313: for ( rank = 0, j = 0; j < col; j++ ) {
5314: for ( i = rank; i < row; i++ )
5315: if ( mat[i][j] )
5316: break;
5317: if ( i == row ) {
5318: colstat[j] = 0;
5319: continue;
5320: } else
5321: colstat[j] = 1;
5322: if ( i != rank ) {
5323: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5324: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5325: }
5326: pivot = mat[rank];
5327: s = sugar[rank];
5328: inv = _invsf(pivot[j]);
5329: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5330: if ( *pk )
5331: *pk = _mulsf(*pk,inv);
5332: for ( i = rank+1; i < row; i++ ) {
5333: t = mat[i];
5334: if ( a = t[j] ) {
5335: sugar[i] = MAX(sugar[i],s);
5336: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5337: }
5338: }
5339: rank++;
5340: }
5341: for ( j = col-1, l = rank-1; j >= 0; j-- )
5342: if ( colstat[j] ) {
5343: pivot = mat[l];
5344: s = sugar[l];
5345: for ( i = 0; i < l; i++ ) {
5346: t = mat[i];
5347: if ( a = t[j] ) {
5348: sugar[i] = MAX(sugar[i],s);
5349: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5350: }
5351: }
5352: l--;
5353: }
5354: return rank;
1.77 noro 5355: }
5356:
5357: int ndv_ishomo(NDV p)
5358: {
5359: NMV m;
5360: int len,h;
5361:
5362: if ( !p ) return 1;
5363: len = LEN(p);
5364: m = BDY(p);
5365: h = TD(DL(m));
5366: NMV_ADV(m);
5367: for ( len--; len; len--, NMV_ADV(m) )
5368: if ( TD(DL(m)) != h ) return 0;
5369: return 1;
5370: }
5371:
5372: void ndv_save(NDV p,int index)
5373: {
5374: FILE *s;
5375: char name[BUFSIZ];
5376: short id;
5377: int nv,sugar,len,n,i,td,e,j;
5378: NMV m;
5379: unsigned int *dl;
5380:
5381: sprintf(name,"%s/%d",Demand,index);
5382: s = fopen(name,"w");
5383: savevl(s,0);
5384: if ( !p ) {
5385: saveobj(s,0);
5386: return;
5387: }
5388: id = O_DP;
5389: nv = NV(p);
5390: sugar = SG(p);
5391: len = LEN(p);
5392: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5393: write_int(s,&len);
5394:
5395: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5396: saveobj(s,(Obj)CQ(m));
1.77 noro 5397: dl = DL(m);
5398: td = TD(dl);
5399: write_int(s,&td);
5400: for ( j = 0; j < nv; j++ ) {
5401: e = GET_EXP(dl,j);
5402: write_int(s,&e);
5403: }
5404: }
5405: fclose(s);
5406: }
5407:
5408: NDV ndv_load(int index)
5409: {
5410: FILE *s;
5411: char name[BUFSIZ];
5412: short id;
5413: int nv,sugar,len,n,i,td,e,j;
5414: NDV d;
5415: NMV m0,m;
5416: unsigned int *dl;
5417: Obj obj;
5418:
5419: sprintf(name,"%s/%d",Demand,index);
5420: s = fopen(name,"r");
5421: if ( !s ) return 0;
5422:
5423: skipvl(s);
5424: read_short(s,&id);
5425: if ( !id ) return 0;
5426: read_int(s,&nv);
5427: read_int(s,&sugar);
5428: read_int(s,&len);
5429:
5430: m0 = m = MALLOC(len*nmv_adv);
5431: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5432: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5433: dl = DL(m);
5434: ndl_zero(dl);
5435: read_int(s,&td); TD(dl) = td;
5436: for ( j = 0; j < nv; j++ ) {
5437: read_int(s,&e);
5438: PUT_EXP(dl,j,e);
5439: }
5440: if ( nd_blockmask ) ndl_weight_mask(dl);
5441: }
5442: fclose(s);
5443: MKNDV(nv,m0,len,d);
5444: SG(d) = sugar;
5445: return d;
1.99 noro 5446: }
5447:
1.102 noro 5448: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5449: {
5450: VL fv,tv;
1.102 noro 5451: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5452: pointer **m;
1.113 noro 5453: Q mone;
1.125 noro 5454: P **w;
1.126 noro 5455: P mp,r;
1.102 noro 5456: NDV **dm;
5457: NDV *t,*mi,*mj;
5458: NDV d,s,mij,mjj;
5459: ND u;
5460: NMV nmv;
1.114 noro 5461: UINT *bound;
1.99 noro 5462: PGeoBucket bucket;
5463: struct order_spec *ord;
1.126 noro 5464: Q dq,dt,ds;
5465: N gn,qn,dn0,nm,dn;
1.99 noro 5466:
5467: create_order_spec(0,0,&ord);
5468: nd_init_ord(ord);
5469: get_vars((Obj)f,&fv);
5470: if ( f->row != f->col )
5471: error("nd_det : non-square matrix");
5472: n = f->row;
1.125 noro 5473: m = f->body;
1.99 noro 5474: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5475:
5476: if ( !nvar ) {
5477: if ( !mod )
5478: detp(CO,(P **)m,n,rp);
5479: else {
5480: w = (P **)almat_pointer(n,n);
5481: for ( i = 0; i < n; i++ )
5482: for ( j = 0; j < n; j++ )
5483: ptomp(mod,(P)m[i][j],&w[i][j]);
5484: detmp(CO,mod,w,n,&mp);
5485: mptop(mp,rp);
5486: }
5487: return;
5488: }
1.126 noro 5489:
5490: if ( !mod ) {
5491: w = (P **)almat_pointer(n,n);
5492: dq = ONE;
5493: for ( i = 0; i < n; i++ ) {
5494: dn0 = ONEN;
5495: for ( j = 0; j < n; j++ ) {
5496: if ( !m[i][j] ) continue;
5497: lgp(m[i][j],&nm,&dn);
5498: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5499: }
5500: if ( !UNIN(dn0) ) {
5501: NTOQ(dn0,1,ds);
5502: for ( j = 0; j < n; j++ )
5503: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5504: mulq(dq,ds,&dt); dq = dt;
5505: } else
5506: for ( j = 0; j < n; j++ )
5507: w[i][j] = (P)m[i][j];
5508: }
5509: m = (pointer **)w;
5510: }
1.125 noro 5511:
1.99 noro 5512: for ( i = 0, max = 0; i < n; i++ )
5513: for ( j = 0; j < n; j++ )
5514: for ( tv = fv; tv; tv = NEXT(tv) ) {
5515: e = getdeg(tv->v,(P)m[i][j]);
5516: max = MAX(e,max);
5517: }
1.114 noro 5518: nd_setup_parameters(nvar,max);
1.102 noro 5519: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5520: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5521: for ( j = 0; j < n; j++ ) {
5522: dm[i][j] = ptondv(CO,fv,m[i][j]);
5523: if ( mod ) ndv_mod(mod,dm[i][j]);
5524: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5525: }
5526: d = ptondv(CO,fv,(P)ONE);
5527: if ( mod ) ndv_mod(mod,d);
1.113 noro 5528: chsgnq(ONE,&mone);
1.99 noro 5529: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5530: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5531: for ( i = j; i < n && !dm[i][j]; i++ );
5532: if ( i == n ) {
5533: *rp = 0;
5534: return;
5535: }
1.102 noro 5536: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5537: for ( k = j; k < n; k++ )
5538: for ( l = j; l < n; l++ )
1.102 noro 5539: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5540: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5541: }
5542: if ( k0 != j ) {
5543: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5544: sgn = -sgn;
5545: }
5546: if ( l0 != j ) {
5547: for ( k = j; k < n; k++ ) {
5548: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5549: }
5550: sgn = -sgn;
5551: }
1.114 noro 5552: bound = nd_det_compute_bound(dm,n,j);
5553: if ( ndl_check_bound(bound,bound) )
5554: nd_det_reconstruct(dm,n,j,d);
5555:
1.99 noro 5556: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5557: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5558: mi = dm[i]; mij = mi[j];
1.102 noro 5559: if ( mod )
5560: ndv_mul_c(mod,mij,mod-1);
5561: else
1.113 noro 5562: ndv_mul_c_q(mij,mone);
1.99 noro 5563: for ( k = j+1; k < n; k++ ) {
1.114 noro 5564: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5565: bucket = create_pbucket();
1.104 noro 5566: if ( mi[k] ) {
1.102 noro 5567: nmv = BDY(mjj); len = LEN(mjj);
5568: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5569: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5570: add_pbucket(mod,bucket,u);
1.99 noro 5571: }
1.104 noro 5572: }
1.99 noro 5573: if ( mj[k] && mij ) {
1.102 noro 5574: nmv = BDY(mij); len = LEN(mij);
5575: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5576: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5577: add_pbucket(mod,bucket,u);
1.99 noro 5578: }
5579: }
1.104 noro 5580: u = nd_quo(mod,bucket,d);
1.102 noro 5581: mi[k] = ndtondv(mod,u);
1.99 noro 5582: }
1.114 noro 5583: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5584: }
5585: d = mjj;
5586: }
1.125 noro 5587: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5588: if ( sgn < 0 )
1.102 noro 5589: if ( mod )
5590: ndv_mul_c(mod,d,mod-1);
5591: else
1.113 noro 5592: ndv_mul_c_q(d,mone);
1.126 noro 5593: r = ndvtop(mod,CO,fv,d);
5594: if ( !mod && !UNIQ(dq) )
5595: divsp(CO,r,(P)dq,rp);
5596: else
5597: *rp = r;
1.99 noro 5598: }
5599:
1.102 noro 5600: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5601: {
5602: NM mr,mr0;
1.102 noro 5603: NM tnm;
5604: NMV m;
1.99 noro 5605: UINT *d0,*dt,*dm;
5606: int c,n,td,i,c1,c2,len;
1.113 noro 5607: Q q;
1.99 noro 5608: ND r;
5609:
5610: if ( !p ) return 0;
5611: else {
1.102 noro 5612: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5613: d0 = DL(m0);
1.102 noro 5614: td = TD(d);
1.99 noro 5615: mr0 = 0;
5616: NEWNM(tnm);
1.102 noro 5617: if ( mod ) {
5618: c = CM(m0);
5619: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5620: ndl_add(DL(m),d0,DL(tnm));
5621: if ( ndl_reducible(DL(tnm),d) ) {
5622: NEXTNM(mr0,mr);
5623: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5624: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5625: }
5626: }
5627: } else {
1.113 noro 5628: q = CQ(m0);
1.102 noro 5629: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5630: ndl_add(DL(m),d0,DL(tnm));
5631: if ( ndl_reducible(DL(tnm),d) ) {
5632: NEXTNM(mr0,mr);
1.113 noro 5633: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5634: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5635: }
1.99 noro 5636: }
5637: }
5638: if ( !mr0 )
5639: return 0;
5640: else {
5641: NEXT(mr) = 0;
1.105 noro 5642: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5643: MKND(NV(p),mr0,len,r);
5644: SG(r) = SG(p) + TD(d0);
5645: return r;
5646: }
5647: }
1.114 noro 5648: }
5649:
5650: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5651: {
5652: int i,obpe,oadv,h,k,l;
5653: static NM prev_nm_free_list;
5654: EPOS oepos;
5655:
5656: obpe = nd_bpe;
5657: oadv = nmv_adv;
5658: oepos = nd_epos;
5659: if ( obpe < 2 ) nd_bpe = 2;
5660: else if ( obpe < 3 ) nd_bpe = 3;
5661: else if ( obpe < 4 ) nd_bpe = 4;
5662: else if ( obpe < 5 ) nd_bpe = 5;
5663: else if ( obpe < 6 ) nd_bpe = 6;
5664: else if ( obpe < 8 ) nd_bpe = 8;
5665: else if ( obpe < 10 ) nd_bpe = 10;
5666: else if ( obpe < 16 ) nd_bpe = 16;
5667: else if ( obpe < 32 ) nd_bpe = 32;
5668: else error("nd_det_reconstruct : exponent too large");
5669:
5670: nd_setup_parameters(nd_nvar,0);
5671: prev_nm_free_list = _nm_free_list;
5672: _nm_free_list = 0;
5673: for ( k = j; k < n; k++ )
5674: for (l = j; l < n; l++ )
5675: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5676: ndv_realloc(d,obpe,oadv,oepos);
5677: prev_nm_free_list = 0;
5678: #if 0
5679: GC_gcollect();
5680: #endif
5681: }
5682:
5683: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5684: {
5685: UINT *d0,*d1,*d,*t,*r;
5686: int k,l;
5687:
5688: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5689: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5690: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5691: for ( k = j; k < n; k++ )
5692: for ( l = j; l < n; l++ )
5693: if ( dm[k][l] ) {
5694: d = ndv_compute_bound(dm[k][l]);
5695: ndl_lcm(d,d0,d1);
5696: t = d1; d1 = d0; d0 = t;
5697: }
5698: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5699: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5700: return r;
1.117 noro 5701: }
5702:
5703: DL nd_separate_d(UINT *d,UINT *trans)
5704: {
1.121 noro 5705: int n,td,i,e,j;
1.117 noro 5706: DL a;
5707:
5708: ndl_zero(trans);
5709: td = 0;
1.121 noro 5710: for ( i = 0; i < nd_ntrans; i++ ) {
5711: e = GET_EXP(d,i);
5712: PUT_EXP(trans,i,e);
5713: td += MUL_WEIGHT(e,i);
5714: }
5715: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5716: /* homogenized */
5717: i = nd_nvar-1;
1.117 noro 5718: e = GET_EXP(d,i);
5719: PUT_EXP(trans,i,e);
5720: td += MUL_WEIGHT(e,i);
5721: }
5722: TD(trans) = td;
5723: if ( nd_blockmask) ndl_weight_mask(trans);
5724: NEWDL(a,nd_nalg);
5725: td = 0;
1.121 noro 5726: for ( i = 0; i < nd_nalg; i++ ) {
5727: j = nd_ntrans+i;
5728: e = GET_EXP(d,j);
5729: a->d[i] = e;
1.117 noro 5730: td += e;
5731: }
5732: a->td = td;
5733: return a;
5734: }
5735:
1.118 noro 5736: int nd_monic(int mod,ND *p)
1.117 noro 5737: {
5738: UINT *trans,*t;
5739: DL alg;
5740: MP mp0,mp;
1.118 noro 5741: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5742: ND r;
1.117 noro 5743: DL dl;
5744: DP nm;
5745: NDV ndv;
1.118 noro 5746: DAlg inv,cd;
1.117 noro 5747: ND s,c;
1.118 noro 5748: Q l,mul;
5749: N ln;
5750: int n,ntrans,i,e,td,is_lc,len;
5751: NumberField nf;
5752: struct oEGT eg0,eg1;
5753:
5754: if ( !(nf = get_numberfield()) )
5755: error("nd_monic : current_numberfield is not set");
1.117 noro 5756:
1.118 noro 5757: /* Q coef -> DAlg coef */
5758: NEWNM(ma0); ma = ma0;
5759: m = BDY(*p);
5760: is_lc = 1;
5761: while ( 1 ) {
5762: NEWMP(mp0); mp = mp0;
5763: mp->c = (P)CQ(m);
5764: mp->dl = nd_separate_d(DL(m),DL(ma));
5765: NEWNM(mb);
5766: for ( m = NEXT(m); m; m = NEXT(m) ) {
5767: alg = nd_separate_d(DL(m),DL(mb));
5768: if ( !ndl_equal(DL(ma),DL(mb)) )
5769: break;
5770: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5771: }
5772: NEXT(mp) = 0;
5773: MKDP(nd_nalg,mp0,nm);
5774: MKDAlg(nm,ONE,cd);
5775: if ( is_lc == 1 ) {
5776: /* if the lc is a rational number, we have nothing to do */
5777: if ( !mp0->dl->td )
5778: return 1;
5779:
5780: get_eg(&eg0);
5781: invdalg(cd,&inv);
5782: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5783: /* check the validity of inv */
5784: if ( mod && !rem(NM(inv->dn),mod) )
5785: return 0;
5786: CA(ma) = nf->one;
5787: is_lc = 0;
5788: ln = ONEN;
5789: } else {
5790: muldalg(cd,inv,&CA(ma));
5791: lcmn(ln,NM(CA(ma)->dn),&ln);
5792: }
5793: if ( m ) {
5794: NEXT(ma) = mb; ma = mb;
5795: } else {
5796: NEXT(ma) = 0;
5797: break;
5798: }
5799: }
5800: /* l = lcm(denoms) */
5801: NTOQ(ln,1,l);
5802: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5803: divq(l,CA(m)->dn,&mul);
5804: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5805: NEXTNM(mr0,mr);
5806: mulq((Q)mp->c,mul,&CQ(mr));
5807: dl = mp->dl;
5808: td = TD(DL(m));
5809: ndl_copy(DL(m),DL(mr));
1.121 noro 5810: for ( i = 0; i < nd_nalg; i++ ) {
5811: e = dl->d[i];
5812: PUT_EXP(DL(mr),i+nd_ntrans,e);
5813: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5814: }
5815: TD(DL(mr)) = td;
5816: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5817: }
1.117 noro 5818: }
1.118 noro 5819: NEXT(mr) = 0;
5820: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5821: MKND(NV(*p),mr0,len,r);
5822: /* XXX */
5823: SG(r) = SG(*p);
5824: nd_free(*p);
5825: *p = r;
5826: return 1;
1.59 noro 5827: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>