Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.145
1.145 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.144 2006/08/26 03:09:55 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:
1.143 noro 1551: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.92 noro 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;
1.140 noro 1595: int diag_count = 0;
1.1 noro 1596:
1.23 noro 1597: g = 0; d = 0;
1598: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1599: d = update_pairs(d,g,i);
1600: g = update_base(g,i);
1601: }
1602: sugar = 0;
1603: while ( d ) {
1604: again:
1605: l = nd_minp(d,&d);
1.14 noro 1606: if ( SG(l) != sugar ) {
1.122 noro 1607: if ( ishomo ) {
1.140 noro 1608: diag_count = 0;
1.122 noro 1609: stat = do_diagonalize(sugar,m);
1610: if ( !stat ) {
1611: NEXT(l) = d; d = l;
1612: d = nd_reconstruct(0,d);
1613: goto again;
1614: }
1615: }
1.14 noro 1616: sugar = SG(l);
1.71 noro 1617: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1618: }
1.53 noro 1619: stat = nd_sp(m,0,l,&h);
1.1 noro 1620: if ( !stat ) {
1621: NEXT(l) = d; d = l;
1.103 noro 1622: d = nd_reconstruct(0,d);
1.1 noro 1623: goto again;
1624: }
1.41 noro 1625: #if USE_GEOBUCKET
1.69 noro 1626: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1627: #else
1.69 noro 1628: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1629: #endif
1.1 noro 1630: if ( !stat ) {
1631: NEXT(l) = d; d = l;
1.103 noro 1632: d = nd_reconstruct(0,d);
1.1 noro 1633: goto again;
1634: } else if ( nf ) {
1.27 noro 1635: if ( checkonly ) return 0;
1.71 noro 1636: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1637: nd_removecont(m,nf);
1.118 noro 1638: if ( !m && nd_nalg ) {
1639: nd_monic(0,&nf);
1640: nd_removecont(m,nf);
1.117 noro 1641: }
1.63 noro 1642: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1643: nh = ndv_newps(m,nfv,0);
1.141 noro 1644: if ( !m && (ishomo && ++diag_count == diag_period) ) {
1.140 noro 1645: diag_count = 0;
1646: stat = do_diagonalize(sugar,m);
1647: if ( !stat ) {
1648: NEXT(l) = d; d = l;
1649: d = nd_reconstruct(1,d);
1650: goto again;
1651: }
1652: }
1.1 noro 1653: d = update_pairs(d,g,nh);
1654: g = update_base(g,nh);
1655: FREENDP(l);
1656: } else {
1.71 noro 1657: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1658: FREENDP(l);
1659: }
1660: }
1.77 noro 1661: if ( nd_demand )
1662: for ( t = g; t; t = NEXT(t) )
1663: BDY(t) = (pointer)ndv_load((int)BDY(t));
1664: else
1665: for ( t = g; t; t = NEXT(t) )
1666: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1667: return g;
1668: }
1669:
1.122 noro 1670: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1671: {
1672: int i,nh,stat;
1673: NODE r,g,t;
1674: ND h,nf,nfq,s,head;
1675: NDV nfv,nfqv;
1.113 noro 1676: Q q,den,num;
1677: union oNDC dn;
1.91 noro 1678:
1.143 noro 1679: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.91 noro 1680: /* for nd_ps */
1681: s = ndvtond(m,nd_ps[i]);
1682: s = nd_separate_head(s,&head);
1.122 noro 1683: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1684: if ( !stat ) return 0;
1.91 noro 1685: nf = nd_add(m,head,nf);
1686: ndv_free(nd_ps[i]);
1687: nd_ps[i] = ndtondv(m,nf);
1688: nd_free(nf);
1689:
1690: /* for nd_ps_trace */
1691: if ( nd_demand )
1692: nfv = ndv_load(i);
1693: else
1694: nfv = nd_ps_trace[i];
1695: s = ndvtond(0,nfv);
1696: s = nd_separate_head(s,&head);
1.122 noro 1697: stat = nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1698: if ( !stat ) return 0;
1.113 noro 1699: NTOQ(NM(dn.z),SGN(dn.z),num);
1700: mulq(HCQ(head),num,&q); HCQ(head) = q;
1701: if ( DN(dn.z) ) {
1702: NTOQ(DN(dn.z),1,den);
1703: nd_mul_c_q(nf,den);
1704: }
1.91 noro 1705: nf = nd_add(0,head,nf);
1706: ndv_free(nfv);
1707: nd_removecont(0,nf);
1708: nfv = ndtondv(0,nf);
1709: nd_free(nf);
1710: nd_bound[i] = ndv_compute_bound(nfv);
1711: if ( nd_demand ) {
1712: ndv_save(nfv,i);
1713: ndv_free(nfv);
1714: } else
1715: nd_ps_trace[i] = nfv;
1716: }
1.122 noro 1717: return 1;
1.91 noro 1718: }
1719:
1.118 noro 1720: static struct oEGT eg_invdalg;
1721: struct oEGT eg_le;
1722:
1.82 noro 1723: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1724: {
1725: int i,nh,sugar,stat;
1.23 noro 1726: NODE r,g,t;
1.20 noro 1727: ND_pairs d;
1728: ND_pairs l;
1.82 noro 1729: ND h,nf,nfq,s,head;
1.63 noro 1730: NDV nfv,nfqv;
1.113 noro 1731: Q q,den,num;
1732: union oNDC dn;
1.118 noro 1733: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1734: int diag_count = 0;
1.20 noro 1735:
1.118 noro 1736: init_eg(&eg_monic);
1737: init_eg(&eg_invdalg);
1738: init_eg(&eg_le);
1.23 noro 1739: g = 0; d = 0;
1740: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1741: d = update_pairs(d,g,i);
1742: g = update_base(g,i);
1743: }
1744: sugar = 0;
1745: while ( d ) {
1746: again:
1747: l = nd_minp(d,&d);
1748: if ( SG(l) != sugar ) {
1.130 noro 1749: #if 1
1.122 noro 1750: if ( ishomo ) {
1751: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1752: diag_count = 0;
1.122 noro 1753: if ( !stat ) {
1754: NEXT(l) = d; d = l;
1755: d = nd_reconstruct(1,d);
1756: goto again;
1757: }
1758: }
1.130 noro 1759: #endif
1.20 noro 1760: sugar = SG(l);
1.71 noro 1761: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1762: }
1.53 noro 1763: stat = nd_sp(m,0,l,&h);
1.20 noro 1764: if ( !stat ) {
1765: NEXT(l) = d; d = l;
1.103 noro 1766: d = nd_reconstruct(1,d);
1.20 noro 1767: goto again;
1768: }
1.41 noro 1769: #if USE_GEOBUCKET
1.53 noro 1770: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1771: #else
1.69 noro 1772: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1773: #endif
1.20 noro 1774: if ( !stat ) {
1775: NEXT(l) = d; d = l;
1.103 noro 1776: d = nd_reconstruct(1,d);
1.20 noro 1777: goto again;
1778: } else if ( nf ) {
1.77 noro 1779: if ( nd_demand ) {
1780: nfqv = ndv_load(nd_psn);
1781: nfq = ndvtond(0,nfqv);
1782: } else
1783: nfq = 0;
1784: if ( !nfq ) {
1785: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1786: NEXT(l) = d; d = l;
1.103 noro 1787: d = nd_reconstruct(1,d);
1.77 noro 1788: goto again;
1789: }
1.70 noro 1790: }
1.20 noro 1791: if ( nfq ) {
1.70 noro 1792: /* m|HC(nfq) => failure */
1.113 noro 1793: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1794:
1.71 noro 1795: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1796: if ( nd_nalg ) {
1797: /* m|DN(HC(nf)^(-1)) => failure */
1798: get_eg(&egm0);
1799: if ( !nd_monic(m,&nfq) ) return 0;
1800: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1801: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1802: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1803: } else {
1804: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1805: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1806: }
1.77 noro 1807: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1808: if ( ishomo && ++diag_count == diag_period ) {
1809: diag_count = 0;
1810: stat = do_diagonalize_trace(sugar,m);
1811: if ( !stat ) {
1812: NEXT(l) = d; d = l;
1813: d = nd_reconstruct(1,d);
1814: goto again;
1815: }
1816: }
1.20 noro 1817: d = update_pairs(d,g,nh);
1818: g = update_base(g,nh);
1819: } else {
1.71 noro 1820: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1821: }
1822: } else {
1.71 noro 1823: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1824: }
1825: FREENDP(l);
1826: }
1.77 noro 1827: if ( nd_demand )
1828: for ( t = g; t; t = NEXT(t) )
1829: BDY(t) = (pointer)ndv_load((int)BDY(t));
1830: else
1831: for ( t = g; t; t = NEXT(t) )
1832: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1833: if ( nd_nalg ) {
1834: print_eg("monic",&eg_monic);
1835: print_eg("invdalg",&eg_invdalg);
1836: print_eg("le",&eg_le);
1837: }
1.20 noro 1838: return g;
1839: }
1840:
1.23 noro 1841: int ndv_compare(NDV *p1,NDV *p2)
1842: {
1.34 noro 1843: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1844: }
1845:
1846: int ndv_compare_rev(NDV *p1,NDV *p2)
1847: {
1.34 noro 1848: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1849: }
1850:
1.61 noro 1851: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1852: {
1.69 noro 1853: int i,n,stat;
1854: ND nf,g,head;
1.23 noro 1855: NODE t,a0,a;
1.113 noro 1856: union oNDC dn;
1.69 noro 1857: NDV *w;
1.113 noro 1858: Q q,num,den;
1.23 noro 1859:
1.69 noro 1860: n = length(f);
1861: #if 0
1862: w = (NDV *)ALLOCA(n*sizeof(NDV));
1863: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1864: qsort(w,n,sizeof(NDV),
1865: (int (*)(const void *,const void *))ndv_compare);
1866: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1867: #endif
1.128 noro 1868: ndv_setup(m,0,f,0);
1.69 noro 1869: for ( i = 0; i < n; ) {
1870: g = ndvtond(m,nd_ps[i]);
1871: g = nd_separate_head(g,&head);
1872: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1873: if ( !stat )
1.103 noro 1874: nd_reconstruct(0,0);
1.69 noro 1875: else {
1.71 noro 1876: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1877: if ( !m ) {
1.113 noro 1878: NTOQ(NM(dn.z),SGN(dn.z),num);
1879: mulq(HCQ(head),num,&q); HCQ(head) = q;
1880: if ( DN(dn.z) ) {
1881: NTOQ(DN(dn.z),1,den);
1882: nd_mul_c_q(nf,den);
1883: }
1.85 noro 1884: }
1.69 noro 1885: nf = nd_add(m,head,nf);
1886: ndv_free(nd_ps[i]);
1.24 noro 1887: nd_removecont(m,nf);
1.69 noro 1888: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1889: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1890: i++;
1.23 noro 1891: }
1892: }
1.71 noro 1893: if ( DP_Print ) { printf("\n"); }
1.23 noro 1894: for ( a0 = 0, i = 0; i < n; i++ ) {
1895: NEXTNODE(a0,a);
1.69 noro 1896: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1897: }
1898: NEXT(a) = 0;
1899: return a0;
1900: }
1901:
1.1 noro 1902: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1903: {
1904: ND_pairs d1,nd,cur,head,prev,remove;
1905:
1906: if ( !g ) return d;
1907: d = crit_B(d,t);
1908: d1 = nd_newpairs(g,t);
1909: d1 = crit_M(d1);
1910: d1 = crit_F(d1);
1.55 noro 1911: if ( do_weyl )
1912: head = d1;
1913: else {
1914: prev = 0; cur = head = d1;
1915: while ( cur ) {
1916: if ( crit_2( cur->i1,cur->i2 ) ) {
1917: remove = cur;
1918: if ( !prev ) head = cur = NEXT(cur);
1919: else cur = NEXT(prev) = NEXT(cur);
1920: FREENDP(remove);
1921: } else {
1922: prev = cur; cur = NEXT(cur);
1923: }
1.1 noro 1924: }
1925: }
1926: if ( !d )
1927: return head;
1928: else {
1929: nd = d;
1.34 noro 1930: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1931: NEXT(nd) = head;
1932: return d;
1933: }
1934: }
1935:
1936: ND_pairs nd_newpairs( NODE g, int t )
1937: {
1938: NODE h;
1.61 noro 1939: UINT *dl;
1.34 noro 1940: int ts,s;
1.1 noro 1941: ND_pairs r,r0;
1942:
1.20 noro 1943: dl = DL(nd_psh[t]);
1.34 noro 1944: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1945: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1946: NEXTND_pairs(r0,r);
1947: r->i1 = (int)BDY(h);
1948: r->i2 = t;
1.20 noro 1949: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1950: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1951: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1952: }
1953: NEXT(r) = 0;
1954: return r0;
1955: }
1956:
1957: ND_pairs crit_B( ND_pairs d, int s )
1958: {
1959: ND_pairs cur,head,prev,remove;
1.61 noro 1960: UINT *t,*tl,*lcm;
1.1 noro 1961: int td,tdl;
1962:
1963: if ( !d ) return 0;
1.20 noro 1964: t = DL(nd_psh[s]);
1.1 noro 1965: prev = 0;
1966: head = cur = d;
1.61 noro 1967: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1968: while ( cur ) {
1969: tl = cur->lcm;
1970: if ( ndl_reducible(tl,t)
1.20 noro 1971: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1972: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1973: remove = cur;
1974: if ( !prev ) {
1975: head = cur = NEXT(cur);
1976: } else {
1977: cur = NEXT(prev) = NEXT(cur);
1978: }
1979: FREENDP(remove);
1980: } else {
1.34 noro 1981: prev = cur; cur = NEXT(cur);
1.1 noro 1982: }
1983: }
1984: return head;
1985: }
1986:
1987: ND_pairs crit_M( ND_pairs d1 )
1988: {
1989: ND_pairs e,d2,d3,dd,p;
1.61 noro 1990: UINT *id,*jd;
1.1 noro 1991:
1992: for ( dd = 0, e = d1; e; e = d3 ) {
1993: if ( !(d2 = NEXT(e)) ) {
1994: NEXT(e) = dd;
1995: return e;
1996: }
1.34 noro 1997: id = LCM(e);
1.1 noro 1998: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1999: p = NEXT(d2);
2000: jd = LCM(d2);
2001: if ( ndl_equal(jd,id) )
2002: ;
2003: else if ( TD(jd) > TD(id) )
1.1 noro 2004: if ( ndl_reducible(jd,id) ) continue;
2005: else ;
1.34 noro 2006: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2007: NEXT(d2) = d3;
2008: d3 = d2;
2009: }
2010: NEXT(e) = dd;
2011: dd = e;
2012: continue;
2013: /**/
2014: delit: NEXT(d2) = d3;
2015: d3 = d2;
2016: for ( ; p; p = d2 ) {
2017: d2 = NEXT(p);
2018: NEXT(p) = d3;
2019: d3 = p;
2020: }
2021: FREENDP(e);
2022: }
2023: return dd;
2024: }
2025:
2026: ND_pairs crit_F( ND_pairs d1 )
2027: {
2028: ND_pairs rest, head,remove;
2029: ND_pairs last, p, r, w;
2030: int s;
2031:
2032: for ( head = last = 0, p = d1; NEXT(p); ) {
2033: r = w = equivalent_pairs(p,&rest);
1.14 noro 2034: s = SG(r);
1.1 noro 2035: w = NEXT(w);
2036: while ( w ) {
2037: if ( crit_2(w->i1,w->i2) ) {
2038: r = w;
2039: w = NEXT(w);
2040: while ( w ) {
2041: remove = w;
2042: w = NEXT(w);
2043: FREENDP(remove);
2044: }
2045: break;
1.14 noro 2046: } else if ( SG(w) < s ) {
1.1 noro 2047: FREENDP(r);
2048: r = w;
1.14 noro 2049: s = SG(r);
1.1 noro 2050: w = NEXT(w);
2051: } else {
2052: remove = w;
2053: w = NEXT(w);
2054: FREENDP(remove);
2055: }
2056: }
2057: if ( last ) NEXT(last) = r;
2058: else head = r;
2059: NEXT(last = r) = 0;
2060: p = rest;
2061: if ( !p ) return head;
2062: }
2063: if ( !last ) return p;
2064: NEXT(last) = p;
2065: return head;
2066: }
2067:
2068: int crit_2( int dp1, int dp2 )
2069: {
1.20 noro 2070: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2071: }
2072:
1.40 noro 2073: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2074: {
2075: ND_pairs w,p,r,s;
1.61 noro 2076: UINT *d;
1.1 noro 2077:
2078: w = d1;
1.34 noro 2079: d = LCM(w);
1.1 noro 2080: s = NEXT(w);
2081: NEXT(w) = 0;
2082: for ( r = 0; s; s = p ) {
2083: p = NEXT(s);
1.34 noro 2084: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2085: NEXT(s) = w; w = s;
1.1 noro 2086: } else {
1.39 noro 2087: NEXT(s) = r; r = s;
1.1 noro 2088: }
2089: }
2090: *prest = r;
2091: return w;
2092: }
2093:
2094: NODE update_base(NODE nd,int ndp)
2095: {
1.61 noro 2096: UINT *dl, *dln;
1.137 noro 2097: NODE last, p, head;
1.1 noro 2098:
1.20 noro 2099: dl = DL(nd_psh[ndp]);
1.1 noro 2100: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2101: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2102: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2103: p = NEXT(p);
2104: if ( last ) NEXT(last) = p;
2105: } else {
2106: if ( !last ) head = p;
2107: p = NEXT(last = p);
2108: }
2109: }
2110: head = append_one(head,ndp);
2111: return head;
2112: }
2113:
2114: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2115: {
2116: ND_pairs m,ml,p,l;
1.61 noro 2117: UINT *lcm;
1.33 noro 2118: int s,td,len,tlen,c,c1;
1.1 noro 2119:
2120: if ( !(p = NEXT(m = d)) ) {
2121: *prest = p;
2122: NEXT(m) = 0;
2123: return m;
2124: }
1.14 noro 2125: s = SG(m);
1.77 noro 2126: if ( !NoSugar ) {
2127: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2128: if ( (SG(p) < s)
2129: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2130: ml = l; m = p; s = SG(m);
2131: }
2132: } else {
2133: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2134: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2135: ml = l; m = p; s = SG(m);
2136: }
2137: }
1.1 noro 2138: if ( !ml ) *prest = NEXT(m);
2139: else {
2140: NEXT(ml) = NEXT(m);
2141: *prest = d;
2142: }
2143: NEXT(m) = 0;
2144: return m;
2145: }
2146:
1.63 noro 2147: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2148: {
1.69 noro 2149: int msugar,i;
1.63 noro 2150: ND_pairs t,dm0,dm,dr0,dr;
2151:
2152: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2153: if ( SG(t) < msugar ) msugar = SG(t);
2154: dm0 = 0; dr0 = 0;
1.69 noro 2155: for ( i = 0, t = d; t; t = NEXT(t) )
2156: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2157: if ( dm0 ) NEXT(dm) = t;
2158: else dm0 = t;
2159: dm = t;
1.69 noro 2160: i++;
1.63 noro 2161: } else {
2162: if ( dr0 ) NEXT(dr) = t;
2163: else dr0 = t;
2164: dr = t;
2165: }
2166: NEXT(dm) = 0;
2167: if ( dr0 ) NEXT(dr) = 0;
2168: *prest = dr0;
2169: return dm0;
2170: }
2171:
1.77 noro 2172: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2173: {
1.3 noro 2174: int len;
1.13 noro 2175: RHist r;
1.20 noro 2176: NDV b;
1.3 noro 2177:
1.1 noro 2178: if ( nd_psn == nd_pslen ) {
2179: nd_pslen *= 2;
1.11 noro 2180: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2181: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2182: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2183: nd_bound = (UINT **)
2184: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2185: }
1.39 noro 2186: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2187: nd_ps[nd_psn] = a;
1.39 noro 2188: if ( aq ) {
1.63 noro 2189: nd_ps_trace[nd_psn] = aq;
2190: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2191: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2192: } else {
1.63 noro 2193: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2194: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2195: }
1.77 noro 2196: if ( nd_demand ) {
2197: if ( aq ) {
2198: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2199: nd_ps_trace[nd_psn] = 0;
2200: } else {
2201: ndv_save(nd_ps[nd_psn],nd_psn);
2202: nd_ps[nd_psn] = 0;
2203: }
2204: }
1.1 noro 2205: return nd_psn++;
2206: }
2207:
1.128 noro 2208: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2209: {
1.5 noro 2210: int i,j,td,len,max;
1.1 noro 2211: NODE s,s0,f0;
1.61 noro 2212: UINT *d;
1.13 noro 2213: RHist r;
1.69 noro 2214: NDV *w;
1.61 noro 2215: NDV a,am;
1.11 noro 2216:
2217: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2218:
1.69 noro 2219: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2220: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2221: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.139 noro 2222: if ( !dont_sort ) {
2223: /* XXX heuristic */
2224: if ( !nd_ord->id && (nd_ord->ord.simple<2) )
2225: qsort(w,nd_psn,sizeof(NDV),
2226: (int (*)(const void *,const void *))ndv_compare_rev);
2227: else
2228: qsort(w,nd_psn,sizeof(NDV),
2229: (int (*)(const void *,const void *))ndv_compare);
2230: }
1.69 noro 2231: nd_pslen = 2*nd_psn;
1.11 noro 2232: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2233: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2234: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2235: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2236:
1.11 noro 2237: if ( !nd_red )
1.13 noro 2238: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2239: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2240: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2241: if ( trace ) {
1.69 noro 2242: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2243: ndv_removecont(0,a);
2244: am = nd_ps[i] = ndv_dup(mod,a);
2245: ndv_mod(mod,am);
2246: ndv_removecont(mod,am);
2247: } else {
1.69 noro 2248: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2249: ndv_removecont(mod,a);
1.39 noro 2250: }
1.61 noro 2251: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2252: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2253: nd_psh[i] = r;
1.77 noro 2254: if ( nd_demand ) {
2255: if ( trace ) {
2256: ndv_save(nd_ps_trace[i],i);
2257: nd_ps_trace[i] = 0;
2258: } else {
2259: ndv_save(nd_ps[i],i);
2260: nd_ps[i] = 0;
2261: }
2262: }
1.20 noro 2263: }
2264: }
2265:
1.119 noro 2266: struct order_spec *append_block(struct order_spec *spec,
2267: int nv,int nalg,int ord);
2268:
1.121 noro 2269: extern VECT current_dl_weight_vector_obj;
2270: static VECT prev_weight_vector_obj;
2271:
1.120 noro 2272: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2273: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2274: {
1.121 noro 2275: NODE alist,t,s,r0,r,arg;
1.120 noro 2276: VL tv;
2277: P poly;
2278: DP d;
2279: Alg alpha,dp;
2280: DAlg inv,da,hc;
2281: MP m;
1.121 noro 2282: int i,nvar,nalg,n;
1.120 noro 2283: NumberField nf;
2284: LIST f1,f2;
2285: struct order_spec *current_spec;
1.121 noro 2286: VECT obj,obj0;
2287: Obj tmp;
1.120 noro 2288:
2289: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2290: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2291:
2292: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2293: NEXTNODE(alist,t); MKV(tv->v,poly);
2294: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2295: tv->v = tv->v->priv;
2296: }
2297: NEXT(t) = 0;
2298:
2299: /* simplification, makeing polynomials monic */
2300: setfield_dalg(alist);
2301: obj_algtodalg(f,&f1);
2302: for ( t = BDY(f); t; t = NEXT(t) ) {
2303: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2304: hc = (DAlg)BDY(d)->c;
2305: if ( NID(hc) == N_DA ) {
2306: invdalg(hc,&inv);
2307: for ( m = BDY(d); m; m = NEXT(m) ) {
2308: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2309: }
2310: }
2311: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2312: }
2313: obj_dalgtoalg(f1,&f);
2314:
2315: /* append alg vars to the var list */
2316: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2317: NEXT(tv) = av;
2318:
2319: /* append a block to ord */
2320: *ord1p = append_block(ord,nvar,nalg,2);
2321:
2322: /* create generator list */
2323: nf = get_numberfield();
2324: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2325: MKAlg(nf->defpoly[i],dp);
2326: MKNODE(s,dp,t); t = s;
2327: }
2328: MKLIST(f1,t);
2329: *alistp = alist;
2330: algobjtorat(f1,f1p);
1.121 noro 2331:
2332: /* creating a new weight vector */
2333: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2334: n = nvar+nalg+1;
2335: MKVECT(obj,n);
2336: if ( obj0 && obj0->len == nvar )
2337: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2338: else
2339: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2340: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2341: BDY(obj)[n-1] = (pointer)ONE;
2342: arg = mknode(1,obj);
2343: Pdp_set_weight(arg,&tmp);
2344: }
2345:
2346: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2347: {
2348: NODE s,t,u0,u;
2349: P p;
2350: VL tv;
2351: Obj obj,tmp;
2352: NODE arg;
2353:
2354: u0 = 0;
2355: for ( t = r; t; t = NEXT(t) ) {
2356: p = (P)BDY(t);
2357: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2358: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2359: }
2360: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2361: NEXTNODE(u0,u);
2362: BDY(u) = (pointer)p;
2363: }
2364: }
2365: arg = mknode(1,prev_weight_vector_obj);
2366: Pdp_set_weight(arg,&tmp);
2367:
2368: return u0;
1.120 noro 2369: }
2370:
1.63 noro 2371: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2372: {
1.119 noro 2373: VL tv,fv,vv,vc,av;
2374: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2375: int e,max,nvar,i;
1.69 noro 2376: NDV b;
1.119 noro 2377: int ishomo,nalg;
2378: Alg alpha,dp;
1.144 noro 2379: P p,zp;
2380: Q dmy;
1.119 noro 2381: LIST f1,f2;
2382: Obj obj;
2383: NumberField nf;
1.120 noro 2384: struct order_spec *ord1;
1.1 noro 2385:
1.78 noro 2386: if ( !m && Demand ) nd_demand = 1;
2387: else nd_demand = 0;
2388:
1.103 noro 2389: #if 0
1.74 noro 2390: ndv_alloc = 0;
1.103 noro 2391: #endif
1.1 noro 2392: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2393: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2394: switch ( ord->id ) {
2395: case 1:
2396: if ( ord->nv != nvar )
2397: error("nd_{gr,f4} : invalid order specification");
2398: break;
2399: default:
2400: break;
2401: }
1.119 noro 2402: nd_nalg = 0;
1.123 noro 2403: av = 0;
1.119 noro 2404: if ( !m ) {
2405: get_algtree((Obj)f,&av);
2406: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2407: nd_ntrans = nvar;
1.119 noro 2408: nd_nalg = nalg;
2409: /* #i -> t#i */
2410: if ( nalg ) {
1.120 noro 2411: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2412: ord = ord1;
2413: f = f1;
1.119 noro 2414: }
1.120 noro 2415: nvar += nalg;
1.119 noro 2416: }
1.32 noro 2417: nd_init_ord(ord);
1.61 noro 2418: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2419: for ( tv = vv; tv; tv = NEXT(tv) ) {
2420: e = getdeg(tv->v,(P)BDY(t));
2421: max = MAX(e,max);
2422: }
2423: nd_setup_parameters(nvar,max);
1.77 noro 2424: ishomo = 1;
1.1 noro 2425: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2426: ptozp((P)BDY(t),1,&dmy,&zp);
2427: b = (pointer)ptondv(CO,vv,zp);
1.77 noro 2428: if ( ishomo )
2429: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2430: if ( m ) ndv_mod(m,b);
1.61 noro 2431: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2432: }
2433: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2434: ndv_setup(m,0,fd0,0);
1.77 noro 2435: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2436: nd_demand = 0;
1.61 noro 2437: x = ndv_reducebase(x);
2438: x = ndv_reduceall(m,x);
1.23 noro 2439: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2440: NEXTNODE(r0,r);
1.61 noro 2441: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2442: }
2443: if ( r0 ) NEXT(r) = 0;
1.121 noro 2444: if ( nalg )
2445: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2446: MKLIST(*rp,r0);
1.103 noro 2447: #if 0
1.74 noro 2448: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2449: #endif
1.127 noro 2450: }
2451:
2452: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2453: {
2454: VL tv,fv,vv,vc,av;
2455: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2456: int e,max,nvar,i;
2457: NDV b;
2458: int ishomo,nalg;
2459: Alg alpha,dp;
1.144 noro 2460: P p,zp;
2461: Q dmy;
1.127 noro 2462: LIST f1,f2;
2463: Obj obj;
2464: NumberField nf;
2465: struct order_spec *ord1;
2466:
2467: get_vars((Obj)f,&fv); pltovl(v,&vv);
2468: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2469: switch ( ord->id ) {
2470: case 1:
2471: if ( ord->nv != nvar )
2472: error("nd_check : invalid order specification");
2473: break;
2474: default:
2475: break;
2476: }
2477: nd_nalg = 0;
2478: av = 0;
2479: if ( !m ) {
2480: get_algtree((Obj)f,&av);
2481: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2482: nd_ntrans = nvar;
2483: nd_nalg = nalg;
2484: /* #i -> t#i */
2485: if ( nalg ) {
2486: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2487: ord = ord1;
2488: f = f1;
2489: }
2490: nvar += nalg;
2491: }
2492: nd_init_ord(ord);
2493: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2494: for ( tv = vv; tv; tv = NEXT(tv) ) {
2495: e = getdeg(tv->v,(P)BDY(t));
2496: max = MAX(e,max);
2497: }
2498: nd_setup_parameters(nvar,max);
2499: ishomo = 1;
2500: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2501: ptozp((P)BDY(t),1,&dmy,&zp);
2502: b = (pointer)ptondv(CO,vv,zp);
1.127 noro 2503: if ( ishomo )
2504: ishomo = ishomo && ndv_ishomo(b);
2505: if ( m ) ndv_mod(m,b);
2506: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2507: }
2508: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2509: ndv_setup(m,0,fd0,0);
1.127 noro 2510: for ( x = 0, i = 0; i < nd_psn; i++ )
2511: x = update_base(x,i);
2512: if ( do_check ) {
2513: x = nd_gb(m,ishomo,1);
2514: if ( !x ) {
2515: *rp = 0;
2516: return;
2517: }
2518: } else {
2519: for ( t = x; t; t = NEXT(t) )
2520: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2521: }
2522: x = ndv_reducebase(x);
2523: x = ndv_reduceall(m,x);
2524: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2525: NEXTNODE(r0,r);
2526: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2527: }
2528: if ( r0 ) NEXT(r) = 0;
2529: if ( nalg )
2530: r0 = postprocess_algcoef(av,alist,r0);
2531: MKLIST(*rp,r0);
1.20 noro 2532: }
2533:
1.133 noro 2534: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2535: {
1.119 noro 2536: VL tv,fv,vv,vc,av;
2537: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2538: int m,nocheck,nvar,mindex,e,max;
2539: NDV c;
2540: NMV a;
1.144 noro 2541: P p,zp;
2542: Q dmy;
1.61 noro 2543: EPOS oepos;
1.119 noro 2544: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2545: Alg alpha,dp;
2546: P poly;
2547: LIST f1,f2;
2548: Obj obj;
2549: NumberField nf;
1.120 noro 2550: struct order_spec *ord1;
1.145 ! noro 2551: struct oEGT eg_check,eg0,eg1;
1.20 noro 2552:
2553: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2554: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2555: switch ( ord->id ) {
2556: case 1:
2557: if ( ord->nv != nvar )
2558: error("nd_gr_trace : invalid order specification");
2559: break;
2560: default:
2561: break;
2562: }
1.119 noro 2563:
2564: get_algtree((Obj)f,&av);
2565: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2566: nd_ntrans = nvar;
1.119 noro 2567: nd_nalg = nalg;
2568: /* #i -> t#i */
2569: if ( nalg ) {
1.120 noro 2570: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2571: ord = ord1;
2572: f = f1;
1.119 noro 2573: }
1.120 noro 2574: nvar += nalg;
1.119 noro 2575:
1.52 noro 2576: nocheck = 0;
2577: mindex = 0;
2578:
1.78 noro 2579: if ( Demand ) nd_demand = 1;
2580: else nd_demand = 0;
2581:
1.52 noro 2582: /* setup modulus */
2583: if ( trace < 0 ) {
2584: trace = -trace;
2585: nocheck = 1;
2586: }
2587: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2588: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2589: for ( tv = vv; tv; tv = NEXT(tv) ) {
2590: e = getdeg(tv->v,(P)BDY(t));
2591: max = MAX(e,max);
1.23 noro 2592: }
1.61 noro 2593: nd_init_ord(ord);
2594: nd_setup_parameters(nvar,max);
2595: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2596: ishomo = 1;
1.61 noro 2597: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2598: ptozp((P)BDY(t),1,&dmy,&zp);
2599: c = (pointer)ptondv(CO,vv,zp);
1.82 noro 2600: if ( ishomo )
2601: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2602: if ( c ) {
2603: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2604: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2605: }
1.61 noro 2606: }
2607: if ( in0 ) NEXT(in) = 0;
2608: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2609: if ( !ishomo && homo ) {
1.122 noro 2610: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2611: c = (NDV)BDY(t); len = LEN(c);
2612: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2613: wmax = MAX(TD(DL(a)),wmax);
2614: }
2615: homogenize_order(ord,nvar,&ord1);
1.89 noro 2616: nd_init_ord(ord1);
1.61 noro 2617: nd_setup_parameters(nvar+1,wmax);
2618: for ( t = fd0; t; t = NEXT(t) )
2619: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2620: }
1.52 noro 2621: while ( 1 ) {
1.77 noro 2622: if ( Demand )
2623: nd_demand = 1;
1.128 noro 2624: ndv_setup(m,1,fd0,0);
1.133 noro 2625: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
1.52 noro 2626: if ( !cand ) {
2627: /* failure */
1.61 noro 2628: if ( trace > 1 ) { *rp = 0; return; }
2629: else m = get_lprime(++mindex);
1.52 noro 2630: continue;
2631: }
1.82 noro 2632: if ( !ishomo && homo ) {
1.27 noro 2633: /* dehomogenization */
1.61 noro 2634: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2635: nd_init_ord(ord);
1.61 noro 2636: nd_setup_parameters(nvar,0);
1.27 noro 2637: }
1.77 noro 2638: nd_demand = 0;
1.61 noro 2639: cand = ndv_reducebase(cand);
2640: cand = ndv_reduceall(0,cand);
1.78 noro 2641: cbpe = nd_bpe;
1.61 noro 2642: if ( nocheck )
2643: break;
1.145 ! noro 2644: get_eg(&eg0);
1.61 noro 2645: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2646: /* success */
2647: break;
2648: else if ( trace > 1 ) {
2649: /* failure */
2650: *rp = 0; return;
1.61 noro 2651: } else {
1.52 noro 2652: /* try the next modulus */
2653: m = get_lprime(++mindex);
1.61 noro 2654: /* reset the parameters */
1.82 noro 2655: if ( !ishomo && homo ) {
1.89 noro 2656: nd_init_ord(ord1);
1.61 noro 2657: nd_setup_parameters(nvar+1,wmax);
2658: } else {
2659: nd_init_ord(ord);
2660: nd_setup_parameters(nvar,max);
2661: }
2662: }
1.52 noro 2663: }
1.145 ! noro 2664: get_eg(&eg1); init_eg(&eg_check); add_eg(&eg_check,&eg0,&eg1);
! 2665: if ( DP_Print )
! 2666: fprintf(asir_out,"check=%fsec\n",eg_check.exectime+eg_check.gctime);
1.27 noro 2667: /* dp->p */
1.61 noro 2668: nd_bpe = cbpe;
1.73 noro 2669: nd_setup_parameters(nd_nvar,0);
1.121 noro 2670: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2671: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2672: if ( nalg )
2673: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2674: MKLIST(*rp,cand);
1.1 noro 2675: }
2676:
1.61 noro 2677: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2678: {
1.61 noro 2679: UINT *d;
1.57 noro 2680: int i,j,l,s,ord_l;
1.43 noro 2681: struct order_pair *op;
1.1 noro 2682:
2683: d = dl->d;
1.41 noro 2684: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2685: if ( nd_blockmask ) {
2686: l = nd_blockmask->n;
2687: op = nd_blockmask->order_pair;
2688: for ( j = 0, s = 0; j < l; j++ ) {
2689: ord_l = op[j].length;
1.57 noro 2690: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2691: }
2692: TD(r) = ndl_weight(r);
1.61 noro 2693: ndl_weight_mask(r);
1.43 noro 2694: } else {
1.56 noro 2695: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2696: TD(r) = ndl_weight(r);
2697: }
1.1 noro 2698: }
2699:
1.61 noro 2700: DL ndltodl(int n,UINT *ndl)
1.1 noro 2701: {
2702: DL dl;
2703: int *d;
1.57 noro 2704: int i,j,l,s,ord_l;
1.43 noro 2705: struct order_pair *op;
1.1 noro 2706:
2707: NEWDL(dl,n);
1.34 noro 2708: dl->td = TD(ndl);
1.1 noro 2709: d = dl->d;
1.43 noro 2710: if ( nd_blockmask ) {
2711: l = nd_blockmask->n;
2712: op = nd_blockmask->order_pair;
2713: for ( j = 0, s = 0; j < l; j++ ) {
2714: ord_l = op[j].length;
1.57 noro 2715: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2716: }
2717: } else {
1.56 noro 2718: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2719: }
1.1 noro 2720: return dl;
2721: }
2722:
1.61 noro 2723: void ndl_print(UINT *dl)
1.1 noro 2724: {
2725: int n;
1.57 noro 2726: int i,j,l,ord_l,s,s0;
1.43 noro 2727: struct order_pair *op;
1.1 noro 2728:
2729: n = nd_nvar;
2730: printf("<<");
1.43 noro 2731: if ( nd_blockmask ) {
2732: l = nd_blockmask->n;
2733: op = nd_blockmask->order_pair;
2734: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2735: ord_l = op[j].length;
1.57 noro 2736: for ( i = 0; i < ord_l; i++, s++ )
2737: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2738: }
2739: } else {
1.56 noro 2740: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2741: }
1.1 noro 2742: printf(">>");
2743: }
2744:
2745: void nd_print(ND p)
2746: {
2747: NM m;
2748:
2749: if ( !p )
2750: printf("0\n");
2751: else {
2752: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2753: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2754: else printf("+%d*",CM(m));
1.14 noro 2755: ndl_print(DL(m));
1.1 noro 2756: }
2757: printf("\n");
2758: }
2759: }
2760:
1.113 noro 2761: void nd_print_q(ND p)
1.16 noro 2762: {
2763: NM m;
2764:
2765: if ( !p )
2766: printf("0\n");
2767: else {
2768: for ( m = BDY(p); m; m = NEXT(m) ) {
2769: printf("+");
1.113 noro 2770: printexpr(CO,(Obj)CQ(m));
1.16 noro 2771: printf("*");
2772: ndl_print(DL(m));
2773: }
2774: printf("\n");
2775: }
2776: }
2777:
1.1 noro 2778: void ndp_print(ND_pairs d)
2779: {
2780: ND_pairs t;
2781:
1.34 noro 2782: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2783: printf("\n");
2784: }
2785:
1.20 noro 2786: void nd_removecont(int mod,ND p)
1.16 noro 2787: {
2788: int i,n;
1.113 noro 2789: Q *w;
2790: Q dvr,t;
1.16 noro 2791: NM m;
1.21 noro 2792: struct oVECT v;
1.113 noro 2793: N q,r;
1.16 noro 2794:
1.71 noro 2795: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2796: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2797: else {
2798: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2799: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2800: v.len = n;
2801: v.body = (pointer *)w;
1.113 noro 2802: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2803: removecont_array(w,n);
1.113 noro 2804: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2805: }
2806: }
2807:
1.21 noro 2808: void nd_removecont2(ND p1,ND p2)
2809: {
2810: int i,n1,n2,n;
1.113 noro 2811: Q *w;
2812: Q dvr,t;
1.21 noro 2813: NM m;
2814: struct oVECT v;
1.113 noro 2815: N q,r;
1.21 noro 2816:
2817: if ( !p1 ) {
2818: nd_removecont(0,p2); return;
2819: } else if ( !p2 ) {
2820: nd_removecont(0,p1); return;
2821: }
2822: n1 = nd_length(p1);
2823: n2 = nd_length(p2);
2824: n = n1+n2;
1.113 noro 2825: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2826: v.len = n;
2827: v.body = (pointer *)w;
1.113 noro 2828: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2829: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2830: removecont_array(w,n);
1.113 noro 2831: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2832: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2833: }
2834:
1.20 noro 2835: void ndv_removecont(int mod,NDV p)
1.16 noro 2836: {
2837: int i,len;
1.113 noro 2838: Q *w;
2839: Q dvr,t;
1.16 noro 2840: NMV m;
2841:
1.71 noro 2842: if ( mod == -1 )
2843: ndv_mul_c(mod,p,_invsf(HCM(p)));
2844: else if ( mod )
1.20 noro 2845: ndv_mul_c(mod,p,invm(HCM(p),mod));
2846: else {
2847: len = p->len;
1.113 noro 2848: w = (Q *)ALLOCA(len*sizeof(Q));
2849: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2850: sortbynm(w,len);
2851: qltozl(w,len,&dvr);
2852: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2853: divq(CQ(m),dvr,&t); CQ(m) = t;
2854: }
1.16 noro 2855: }
1.21 noro 2856: }
2857:
1.61 noro 2858: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2859: {
2860: int len,i,max;
2861: NMV m,mr0,mr,t;
2862:
2863: len = p->len;
2864: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2865: max = MAX(max,TD(DL(m)));
2866: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2867: m = (NMV)((char *)mr0+(len-1)*oadv);
2868: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2869: t = (NMV)ALLOCA(nmv_adv);
2870: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2871: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2872: CQ(mr) = CQ(m);
1.61 noro 2873: ndl_copy(DL(t),DL(mr));
2874: }
2875: NV(p)++;
2876: BDY(p) = mr0;
2877: }
2878:
1.45 noro 2879: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2880: {
1.45 noro 2881: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2882: Q *w;
2883: Q dvr,t;
1.23 noro 2884: NMV m,r;
2885:
2886: len = p->len;
2887: newnvar = nd_nvar-1;
1.48 noro 2888: newexporigin = nd_get_exporigin(ord);
1.45 noro 2889: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2890: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2891: ndl_dehomogenize(DL(m));
1.23 noro 2892: if ( newwpd != nd_wpd ) {
1.90 noro 2893: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2894: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2895: CQ(r) = CQ(m);
1.45 noro 2896: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2897: adj = nd_exporigin-newexporigin;
2898: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2899: }
2900: }
2901: NV(p)--;
2902: }
2903:
1.113 noro 2904: void removecont_array(Q *c,int n)
1.21 noro 2905: {
2906: struct oVECT v;
1.113 noro 2907: Q d0,d1,a,u,u1,gcd;
1.110 noro 2908: int i,j;
1.113 noro 2909: N qn,rn,gn;
2910: Q *q,*r;
1.21 noro 2911:
1.113 noro 2912: q = (Q *)ALLOCA(n*sizeof(Q));
2913: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2914: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2915: igcdv_estimate(&v,&d0);
2916: for ( i = 0; i < n; i++ ) {
2917: divn(NM(c[i]),NM(d0),&qn,&rn);
2918: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2919: NTOQ(rn,SGN(c[i]),r[i]);
2920: }
1.34 noro 2921: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2922: if ( i < n ) {
2923: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2924: igcdv(&v,&d1);
2925: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2926: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2927: for ( i = 0; i < n; i++ ) {
1.113 noro 2928: mulq(a,q[i],&u);
2929: if ( r[i] ) {
2930: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2931: addq(u,u1,&q[i]);
2932: } else
2933: q[i] = u;
1.21 noro 2934: }
2935: }
1.34 noro 2936: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2937: }
2938:
1.19 noro 2939: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2940: {
2941: NM m;
2942: int c,c1;
2943:
1.34 noro 2944: if ( !p ) return;
1.115 noro 2945: if ( mul == 1 ) return;
1.71 noro 2946: if ( mod == -1 )
2947: for ( m = BDY(p); m; m = NEXT(m) )
2948: CM(m) = _mulsf(CM(m),mul);
2949: else
2950: for ( m = BDY(p); m; m = NEXT(m) ) {
2951: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2952: }
1.1 noro 2953: }
2954:
1.113 noro 2955: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2956: {
2957: NM m;
1.113 noro 2958: Q c;
1.16 noro 2959:
1.113 noro 2960: if ( !p ) return;
1.115 noro 2961: if ( UNIQ(mul) ) return;
1.16 noro 2962: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2963: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2964: }
2965: }
2966:
1.61 noro 2967: void nd_mul_c_p(VL vl,ND p,P mul)
2968: {
2969: NM m;
2970: P c;
2971:
2972: if ( !p ) return;
2973: for ( m = BDY(p); m; m = NEXT(m) ) {
2974: mulp(vl,CP(m),mul,&c); CP(m) = c;
2975: }
2976: }
2977:
1.1 noro 2978: void nd_free(ND p)
2979: {
2980: NM t,s;
2981:
1.34 noro 2982: if ( !p ) return;
1.1 noro 2983: t = BDY(p);
2984: while ( t ) {
2985: s = NEXT(t);
2986: FREENM(t);
2987: t = s;
2988: }
2989: FREEND(p);
2990: }
2991:
1.23 noro 2992: void ndv_free(NDV p)
2993: {
2994: GC_free(BDY(p));
2995: }
2996:
1.61 noro 2997: void nd_append_red(UINT *d,int i)
1.1 noro 2998: {
1.13 noro 2999: RHist m,m0;
1.1 noro 3000: int h;
3001:
1.13 noro 3002: NEWRHist(m);
1.34 noro 3003: h = ndl_hash_value(d);
1.13 noro 3004: m->index = i;
1.14 noro 3005: ndl_copy(d,DL(m));
1.1 noro 3006: NEXT(m) = nd_red[h];
3007: nd_red[h] = m;
3008: }
3009:
1.61 noro 3010: UINT *ndv_compute_bound(NDV p)
1.1 noro 3011: {
1.61 noro 3012: UINT *d1,*d2,*t;
3013: UINT u;
1.57 noro 3014: int i,j,k,l,len,ind;
1.45 noro 3015: NMV m;
1.1 noro 3016:
3017: if ( !p )
3018: return 0;
1.61 noro 3019: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3020: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 3021: len = LEN(p);
3022: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3023: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3024: ndl_lcm(DL(m),d1,d2);
1.1 noro 3025: t = d1; d1 = d2; d2 = t;
3026: }
1.12 noro 3027: l = nd_nvar+31;
1.61 noro 3028: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3029: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3030: u = d1[i];
3031: k = (nd_epw-1)*nd_bpe;
3032: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3033: t[ind] = (u>>k)&nd_mask0;
3034: }
3035: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3036: return t;
3037: }
3038:
1.99 noro 3039: UINT *nd_compute_bound(ND p)
3040: {
3041: UINT *d1,*d2,*t;
3042: UINT u;
3043: int i,j,k,l,len,ind;
3044: NM m;
3045:
3046: if ( !p )
3047: return 0;
3048: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3049: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3050: len = LEN(p);
3051: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3052: for ( m = NEXT(m); m; m = NEXT(m) ) {
3053: ndl_lcm(DL(m),d1,d2);
3054: t = d1; d1 = d2; d2 = t;
3055: }
3056: l = nd_nvar+31;
3057: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3058: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3059: u = d1[i];
3060: k = (nd_epw-1)*nd_bpe;
3061: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3062: t[ind] = (u>>k)&nd_mask0;
3063: }
3064: for ( ; ind < l; ind++ ) t[ind] = 0;
3065: return t;
3066: }
3067:
1.48 noro 3068: int nd_get_exporigin(struct order_spec *ord)
3069: {
1.51 noro 3070: switch ( ord->id ) {
1.96 noro 3071: case 0: case 2:
1.48 noro 3072: return 1;
1.41 noro 3073: case 1:
3074: /* block order */
1.43 noro 3075: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3076: return ord->ord.block.length+1;
1.96 noro 3077: case 3:
3078: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3079: }
1.48 noro 3080: }
3081:
1.61 noro 3082: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3083: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3084: struct order_pair *op;
1.48 noro 3085:
1.73 noro 3086: nd_nvar = nvar;
1.79 noro 3087: if ( max ) {
1.83 noro 3088: /* XXX */
3089: if ( do_weyl ) nd_bpe = 32;
3090: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3091: else if ( max < 4 ) nd_bpe = 2;
3092: else if ( max < 8 ) nd_bpe = 3;
3093: else if ( max < 16 ) nd_bpe = 4;
3094: else if ( max < 32 ) nd_bpe = 5;
3095: else if ( max < 64 ) nd_bpe = 6;
3096: else if ( max < 256 ) nd_bpe = 8;
3097: else if ( max < 1024 ) nd_bpe = 10;
3098: else if ( max < 65536 ) nd_bpe = 16;
3099: else nd_bpe = 32;
3100: }
1.61 noro 3101: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3102: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3103: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3104: wpd = nd_exporigin+elen;
3105: if ( wpd != nd_wpd ) {
3106: nd_free_private_storage();
3107: nd_wpd = wpd;
3108: }
1.1 noro 3109: if ( nd_bpe < 32 ) {
3110: nd_mask0 = (1<<nd_bpe)-1;
3111: } else {
3112: nd_mask0 = 0xffffffff;
3113: }
3114: bzero(nd_mask,sizeof(nd_mask));
3115: nd_mask1 = 0;
3116: for ( i = 0; i < nd_epw; i++ ) {
3117: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3118: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3119: }
1.90 noro 3120: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3121: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3122: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3123: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3124: }
3125:
1.103 noro 3126: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3127: {
1.37 noro 3128: int i,obpe,oadv,h;
1.71 noro 3129: static NM prev_nm_free_list;
3130: static ND_pairs prev_ndp_free_list;
1.13 noro 3131: RHist mr0,mr;
3132: RHist r;
1.37 noro 3133: RHist *old_red;
1.71 noro 3134: ND_pairs s0,s,t;
1.43 noro 3135: EPOS oepos;
1.15 noro 3136:
1.1 noro 3137: obpe = nd_bpe;
1.11 noro 3138: oadv = nmv_adv;
1.43 noro 3139: oepos = nd_epos;
1.63 noro 3140: if ( obpe < 2 ) nd_bpe = 2;
3141: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3142: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3143: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3144: else if ( obpe < 6 ) nd_bpe = 6;
3145: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3146: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3147: else if ( obpe < 16 ) nd_bpe = 16;
3148: else if ( obpe < 32 ) nd_bpe = 32;
3149: else error("nd_reconstruct : exponent too large");
1.5 noro 3150:
1.73 noro 3151: nd_setup_parameters(nd_nvar,0);
1.1 noro 3152: prev_nm_free_list = _nm_free_list;
3153: prev_ndp_free_list = _ndp_free_list;
3154: _nm_free_list = 0;
3155: _ndp_free_list = 0;
1.53 noro 3156: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3157: if ( trace )
3158: for ( i = nd_psn-1; i >= 0; i-- )
3159: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3160: s0 = 0;
3161: for ( t = d; t; t = NEXT(t) ) {
3162: NEXTND_pairs(s0,s);
3163: s->i1 = t->i1;
3164: s->i2 = t->i2;
1.14 noro 3165: SG(s) = SG(t);
1.61 noro 3166: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3167: }
1.37 noro 3168:
3169: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3170: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3171: old_red[i] = nd_red[i];
3172: nd_red[i] = 0;
3173: }
3174: for ( i = 0; i < REDTAB_LEN; i++ )
3175: for ( r = old_red[i]; r; r = NEXT(r) ) {
3176: NEWRHist(mr);
1.13 noro 3177: mr->index = r->index;
1.20 noro 3178: SG(mr) = SG(r);
1.61 noro 3179: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3180: h = ndl_hash_value(DL(mr));
3181: NEXT(mr) = nd_red[h];
3182: nd_red[h] = mr;
1.6 noro 3183: }
1.37 noro 3184: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3185: old_red = 0;
1.11 noro 3186: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3187: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3188: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3189: nd_psh[i] = r;
1.11 noro 3190: }
1.1 noro 3191: if ( s0 ) NEXT(s) = 0;
3192: prev_nm_free_list = 0;
3193: prev_ndp_free_list = 0;
1.71 noro 3194: #if 0
1.1 noro 3195: GC_gcollect();
1.71 noro 3196: #endif
1.1 noro 3197: return s0;
3198: }
3199:
1.61 noro 3200: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3201: {
1.57 noro 3202: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3203: struct order_pair *op;
1.1 noro 3204:
3205: n = nd_nvar;
1.61 noro 3206: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3207: omask0 = (1<<obpe)-1;
1.34 noro 3208: TD(r) = TD(d);
1.41 noro 3209: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3210: if ( nd_blockmask ) {
3211: l = nd_blockmask->n;
3212: op = nd_blockmask->order_pair;
3213: for ( i = 1; i < nd_exporigin; i++ )
3214: r[i] = d[i];
3215: for ( j = 0, s = 0; j < l; j++ ) {
3216: ord_l = op[j].length;
1.57 noro 3217: for ( i = 0; i < ord_l; i++, s++ ) {
3218: ei = GET_EXP_OLD(d,s);
3219: PUT_EXP(r,s,ei);
3220: }
1.1 noro 3221: }
1.43 noro 3222: } else {
1.56 noro 3223: for ( i = 0; i < n; i++ ) {
3224: ei = GET_EXP_OLD(d,i);
3225: PUT_EXP(r,i,ei);
3226: }
1.1 noro 3227: }
3228: }
1.3 noro 3229:
1.6 noro 3230: ND nd_copy(ND p)
3231: {
3232: NM m,mr,mr0;
1.41 noro 3233: int c,n;
1.6 noro 3234: ND r;
3235:
3236: if ( !p )
3237: return 0;
3238: else {
3239: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3240: NEXTNM(mr0,mr);
1.14 noro 3241: CM(mr) = CM(m);
3242: ndl_copy(DL(m),DL(mr));
1.6 noro 3243: }
3244: NEXT(mr) = 0;
1.31 noro 3245: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3246: SG(r) = SG(p);
1.6 noro 3247: return r;
3248: }
3249: }
3250:
1.53 noro 3251: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3252: {
3253: NM m;
3254: NDV p1,p2;
3255: ND t1,t2;
1.61 noro 3256: UINT *lcm;
1.31 noro 3257: int td;
1.11 noro 3258:
1.77 noro 3259: if ( !mod && nd_demand ) {
3260: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3261: } else {
1.77 noro 3262: if ( trace ) {
3263: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3264: } else {
3265: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3266: }
1.20 noro 3267: }
1.34 noro 3268: lcm = LCM(p);
1.11 noro 3269: NEWNM(m);
1.113 noro 3270: CQ(m) = HCQ(p2);
1.34 noro 3271: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3272: if ( ndl_check_bound2(p->i1,DL(m)) )
3273: return 0;
1.55 noro 3274: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3275: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3276: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3277: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3278: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3279: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3280: nd_free(t1);
3281: return 0;
3282: }
1.55 noro 3283: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3284: *rp = nd_add(mod,t1,t2);
1.11 noro 3285: FREENM(m);
3286: return 1;
3287: }
3288:
1.19 noro 3289: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3290: {
3291: NMV m;
3292: int c,c1,len,i;
3293:
1.34 noro 3294: if ( !p ) return;
1.14 noro 3295: len = LEN(p);
1.71 noro 3296: if ( mod == -1 )
3297: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3298: CM(m) = _mulsf(CM(m),mul);
3299: else
3300: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3301: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3302: }
1.11 noro 3303: }
3304:
1.113 noro 3305: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3306: {
3307: NMV m;
1.113 noro 3308: Q c;
1.16 noro 3309: int len,i;
3310:
1.34 noro 3311: if ( !p ) return;
1.16 noro 3312: len = LEN(p);
3313: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3314: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3315: }
3316: }
3317:
1.55 noro 3318: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3319: int n2,i,j,l,n,tlen;
1.61 noro 3320: UINT *d0;
1.55 noro 3321: NM *tab,*psum;
3322: ND s,r;
3323: NM t;
3324: NMV m1;
3325:
3326: if ( !p ) return 0;
3327: n = NV(p); n2 = n>>1;
3328: d0 = DL(m0);
3329: l = LEN(p);
3330: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3331: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3332: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3333: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3334: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3335: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3336: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3337: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3338: for ( j = 0; j < tlen; j++ ) {
3339: if ( tab[j] ) {
3340: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3341: }
3342: }
3343: }
3344: for ( i = tlen-1, r = 0; i >= 0; i-- )
3345: if ( psum[i] ) {
3346: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3347: MKND(n,psum[i],j,s);
3348: r = nd_add(mod,r,s);
3349: }
1.56 noro 3350: if ( r ) SG(r) = SG(p)+TD(d0);
3351: return r;
1.55 noro 3352: }
3353:
1.56 noro 3354: /* product of monomials */
3355: /* XXX block order is not handled correctly */
3356:
1.55 noro 3357: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3358: {
1.56 noro 3359: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3360: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3361: Q *ctab_q;
3362: Q q,q1;
1.61 noro 3363: UINT c0,c1,c;
1.55 noro 3364: NM *p;
3365: NM m,t;
3366:
3367: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3368: if ( !m0 || !m1 ) return;
3369: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3370: NEWNM(m); d = DL(m);
1.56 noro 3371: if ( mod ) {
3372: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3373: } else
1.113 noro 3374: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3375: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3376: homo = n&1 ? 1 : 0;
3377: if ( homo ) {
3378: /* offset of h-degree */
3379: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3380: PUT_EXP(DL(m),n-1,h);
3381: TD(DL(m)) = h;
1.61 noro 3382: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3383: }
3384: tab[0] = m;
3385: NEWNM(m); d = DL(m);
1.57 noro 3386: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3387: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3388: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3389: /* xi^a*(Di^k*xi^l)*Di^b */
3390: a += l; b += k;
1.56 noro 3391: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3392: if ( !k || !l ) {
3393: for ( j = 0; j < curlen; j++ )
1.56 noro 3394: if ( t = tab[j] ) {
3395: dt = DL(t);
3396: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3397: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3398: }
3399: curlen *= k+1;
3400: continue;
3401: }
3402: min = MIN(k,l);
1.56 noro 3403: if ( mod ) {
1.61 noro 3404: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3405: mkwcm(k,l,mod,ctab);
3406: } else {
1.113 noro 3407: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3408: mkwc(k,l,ctab_q);
1.56 noro 3409: }
1.57 noro 3410: for ( j = min; j >= 0; j-- ) {
1.56 noro 3411: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3412: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3413: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3414: if ( homo ) {
3415: TD(d) = s;
1.56 noro 3416: PUT_EXP(d,n-1,s-h);
1.55 noro 3417: } else TD(d) = h;
1.61 noro 3418: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3419: if ( mod ) c = ctab[j];
1.113 noro 3420: else q = ctab_q[j];
1.57 noro 3421: p = tab+curlen*j;
3422: if ( j == 0 ) {
3423: for ( u = 0; u < curlen; u++, p++ ) {
3424: if ( tab[u] ) {
3425: ndl_addto(DL(tab[u]),d);
3426: if ( mod ) {
3427: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3428: } else {
1.113 noro 3429: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3430: }
3431: }
1.56 noro 3432: }
1.57 noro 3433: } else {
3434: for ( u = 0; u < curlen; u++, p++ ) {
3435: if ( tab[u] ) {
3436: NEWNM(t);
3437: ndl_add(DL(tab[u]),d,DL(t));
3438: if ( mod ) {
3439: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3440: } else
1.113 noro 3441: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3442: *p = t;
3443: }
1.55 noro 3444: }
3445: }
3446: }
3447: curlen *= k+1;
3448: }
3449: FREENM(m);
3450: }
3451:
1.63 noro 3452: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3453: {
3454: NM mr,mr0;
3455: NMV m;
3456: UINT *d,*dt,*dm;
3457: int c,n,td,i,c1,c2,len;
1.113 noro 3458: Q q;
1.63 noro 3459: ND r;
3460:
3461: if ( !p ) return 0;
3462: else {
3463: n = NV(p); m = BDY(p);
3464: d = DL(m0);
3465: len = LEN(p);
3466: mr0 = 0;
3467: td = TD(d);
3468: c = CM(m0);
3469: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3470: NEXTNM(mr0,mr);
3471: CM(mr) = 1;
3472: ndl_add(DL(m),d,DL(mr));
3473: }
3474: NEXT(mr) = 0;
3475: MKND(NV(p),mr0,len,r);
3476: SG(r) = SG(p) + TD(d);
3477: return r;
3478: }
3479: }
3480:
1.55 noro 3481: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3482: {
3483: NM mr,mr0;
3484: NMV m;
1.61 noro 3485: UINT *d,*dt,*dm;
1.9 noro 3486: int c,n,td,i,c1,c2,len;
1.113 noro 3487: Q q;
1.9 noro 3488: ND r;
3489:
1.34 noro 3490: if ( !p ) return 0;
1.55 noro 3491: else if ( do_weyl )
1.71 noro 3492: if ( mod == -1 )
3493: error("ndv_mul_nm : not implemented (weyl)");
3494: else
3495: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3496: else {
3497: n = NV(p); m = BDY(p);
1.34 noro 3498: d = DL(m0);
1.14 noro 3499: len = LEN(p);
1.9 noro 3500: mr0 = 0;
1.34 noro 3501: td = TD(d);
1.71 noro 3502: if ( mod == -1 ) {
3503: c = CM(m0);
3504: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3505: NEXTNM(mr0,mr);
3506: CM(mr) = _mulsf(CM(m),c);
3507: ndl_add(DL(m),d,DL(mr));
3508: }
3509: } else if ( mod ) {
1.16 noro 3510: c = CM(m0);
3511: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3512: NEXTNM(mr0,mr);
3513: c1 = CM(m);
1.19 noro 3514: DMAR(c1,c,0,mod,c2);
1.16 noro 3515: CM(mr) = c2;
3516: ndl_add(DL(m),d,DL(mr));
3517: }
3518: } else {
1.113 noro 3519: q = CQ(m0);
1.16 noro 3520: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3521: NEXTNM(mr0,mr);
1.113 noro 3522: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3523: ndl_add(DL(m),d,DL(mr));
3524: }
1.4 noro 3525: }
1.9 noro 3526: NEXT(mr) = 0;
1.31 noro 3527: MKND(NV(p),mr0,len,r);
1.34 noro 3528: SG(r) = SG(p) + TD(d);
1.9 noro 3529: return r;
1.4 noro 3530: }
3531: }
3532:
1.104 noro 3533: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3534: {
3535: NM mq0,mq;
1.102 noro 3536: NMV tm;
1.113 noro 3537: Q q;
1.104 noro 3538: int i,nv,sg,c,c1,c2,hindex;
3539: ND p,t,r;
3540: N tnm;
3541:
1.124 noro 3542: if ( bucket->m < 0 ) return 0;
1.99 noro 3543: else {
1.104 noro 3544: nv = NV(d);
1.99 noro 3545: mq0 = 0;
1.102 noro 3546: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3547: while ( 1 ) {
1.113 noro 3548: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3549: if ( hindex < 0 ) break;
3550: p = bucket->body[hindex];
1.99 noro 3551: NEXTNM(mq0,mq);
3552: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3553: ndl_copy(DL(mq),DL(tm));
3554: if ( mod ) {
3555: c1 = invm(HCM(d),mod); c2 = HCM(p);
3556: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3557: CM(tm) = mod-c;
3558: } else {
1.113 noro 3559: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3560: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3561: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3562: }
3563: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3564: bucket->body[hindex] = nd_remove_head(p);
3565: t = nd_remove_head(t);
3566: add_pbucket(mod,bucket,t);
3567: }
3568: if ( !mq0 )
3569: r = 0;
3570: else {
3571: NEXT(mq) = 0;
3572: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3573: MKND(nv,mq0,i,r);
3574: /* XXX */
3575: SG(r) = HTD(r);
1.99 noro 3576: }
3577: return r;
3578: }
3579: }
3580:
1.43 noro 3581: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3582: {
1.13 noro 3583: NMV m,mr,mr0,t;
3584: int len,i,k;
1.11 noro 3585:
1.61 noro 3586: if ( !p ) return;
3587: m = BDY(p); len = LEN(p);
3588: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3589: m = (NMV)((char *)mr0+(len-1)*oadv);
3590: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3591: t = (NMV)ALLOCA(nmv_adv);
3592: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3593: CQ(t) = CQ(m);
1.61 noro 3594: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3595: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3596: CQ(mr) = CQ(t);
1.61 noro 3597: ndl_copy(DL(t),DL(mr));
3598: }
3599: BDY(p) = mr0;
3600: }
3601:
3602: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3603: {
3604: NMV m,mr,mr0;
3605: int len,i;
3606: NDV r;
1.11 noro 3607:
1.61 noro 3608: if ( !p ) return 0;
3609: m = BDY(p); len = LEN(p);
3610: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3611: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3612: ndl_zero(DL(mr));
3613: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3614: CQ(mr) = CQ(m);
1.11 noro 3615: }
1.61 noro 3616: MKNDV(NV(p),mr0,len,r);
3617: SG(r) = SG(p);
3618: return r;
1.11 noro 3619: }
3620:
1.61 noro 3621: /* duplicate p */
3622:
3623: NDV ndv_dup(int mod,NDV p)
1.3 noro 3624: {
3625: NDV d;
1.61 noro 3626: NMV t,m,m0;
1.3 noro 3627: int i,len;
3628:
1.34 noro 3629: if ( !p ) return 0;
1.31 noro 3630: len = LEN(p);
1.34 noro 3631: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3632: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3633: ndl_copy(DL(t),DL(m));
1.113 noro 3634: CQ(m) = CQ(t);
1.3 noro 3635: }
3636: MKNDV(NV(p),m0,len,d);
1.23 noro 3637: SG(d) = SG(p);
3638: return d;
3639: }
3640:
1.63 noro 3641: ND nd_dup(ND p)
3642: {
3643: ND d;
3644: NM t,m,m0;
3645:
3646: if ( !p ) return 0;
3647: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3648: NEXTNM(m0,m);
3649: ndl_copy(DL(t),DL(m));
1.113 noro 3650: CQ(m) = CQ(t);
1.63 noro 3651: }
3652: if ( m0 ) NEXT(m) = 0;
3653: MKND(NV(p),m0,LEN(p),d);
3654: SG(d) = SG(p);
3655: return d;
3656: }
3657:
1.61 noro 3658: /* XXX if p->len == 0 then it represents 0 */
3659:
3660: void ndv_mod(int mod,NDV p)
3661: {
3662: NMV t,d;
1.125 noro 3663: int r,s,u;
1.61 noro 3664: int i,len,dlen;
1.71 noro 3665: Obj gfs;
1.61 noro 3666:
3667: if ( !p ) return;
3668: len = LEN(p);
3669: dlen = 0;
1.71 noro 3670: if ( mod == -1 )
3671: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3672: simp_ff((Obj)CP(t),&gfs);
3673: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3674: CM(d) = r;
3675: ndl_copy(DL(t),DL(d));
3676: NMV_ADV(d);
3677: dlen++;
3678: }
1.71 noro 3679: else
3680: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3681: r = rem(NM(CQ(t)),mod);
1.71 noro 3682: if ( r ) {
1.113 noro 3683: if ( SGN(CQ(t)) < 0 )
3684: r = mod-r;
1.125 noro 3685: if ( DN(CQ(t)) ) {
3686: s = rem(DN(CQ(t)),mod);
3687: if ( !s )
3688: error("ndv_mod : division by 0");
3689: s = invm(s,mod);
3690: DMAR(r,s,0,mod,u); r = u;
3691: }
1.71 noro 3692: CM(d) = r;
3693: ndl_copy(DL(t),DL(d));
3694: NMV_ADV(d);
3695: dlen++;
3696: }
3697: }
1.61 noro 3698: LEN(p) = dlen;
3699: }
3700:
3701: NDV ptondv(VL vl,VL dvl,P p)
3702: {
3703: ND nd;
3704:
3705: nd = ptond(vl,dvl,p);
3706: return ndtondv(0,nd);
3707: }
3708:
3709: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3710: {
1.61 noro 3711: int n,i,j,k,e;
3712: VL tvl;
3713: V v;
3714: DCP dc;
3715: DCP *w;
3716: ND r,s,t,u;
3717: P x;
3718: int c;
3719: UINT *d;
1.23 noro 3720: NM m,m0;
1.61 noro 3721:
3722: if ( !p )
3723: return 0;
3724: else if ( NUM(p) ) {
3725: NEWNM(m);
3726: ndl_zero(DL(m));
1.113 noro 3727: CQ(m) = (Q)p;
1.61 noro 3728: NEXT(m) = 0;
3729: MKND(nd_nvar,m,1,r);
3730: SG(r) = 0;
3731: return r;
3732: } else {
3733: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3734: w = (DCP *)ALLOCA(k*sizeof(DCP));
3735: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3736: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3737: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3738: if ( !tvl ) {
3739: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3740: t = ptond(vl,dvl,COEF(w[j]));
3741: pwrp(vl,x,DEG(w[j]),&p);
3742: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3743: }
3744: return s;
3745: } else {
3746: NEWNM(m0); d = DL(m0);
3747: for ( j = k-1, s = 0; j >= 0; j-- ) {
3748: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3749: TD(d) = MUL_WEIGHT(e,i);
3750: if ( nd_blockmask) ndl_weight_mask(d);
3751: t = ptond(vl,dvl,COEF(w[j]));
3752: for ( m = BDY(t); m; m = NEXT(m) )
3753: ndl_addto(DL(m),d);
3754: SG(t) += TD(d);
3755: s = nd_add(0,s,t);
3756: }
3757: FREENM(m0);
3758: return s;
3759: }
3760: }
3761: }
3762:
3763: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3764: {
3765: VL tvl;
3766: int len,n,j,i,e;
3767: NMV m;
3768: Q q;
3769: P c;
3770: UINT *d;
3771: P s,r,u,t,w;
1.71 noro 3772: GFS gfs;
1.23 noro 3773:
1.34 noro 3774: if ( !p ) return 0;
1.61 noro 3775: else {
3776: len = LEN(p);
3777: n = NV(p);
3778: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3779: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3780: if ( mod == -1 ) {
3781: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3782: } else if ( mod ) {
1.61 noro 3783: STOQ(CM(m),q); c = (P)q;
3784: } else
1.113 noro 3785: c = CP(m);
1.61 noro 3786: d = DL(m);
3787: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3788: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3789: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3790: }
3791: addp(vl,s,t,&u); s = u;
3792: }
3793: return s;
1.23 noro 3794: }
1.3 noro 3795: }
3796:
1.61 noro 3797: NDV ndtondv(int mod,ND p)
1.11 noro 3798: {
3799: NDV d;
1.61 noro 3800: NMV m,m0;
3801: NM t;
3802: int i,len;
1.11 noro 3803:
1.34 noro 3804: if ( !p ) return 0;
1.61 noro 3805: len = LEN(p);
1.74 noro 3806: if ( mod )
3807: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3808: else
3809: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3810: #if 0
1.74 noro 3811: ndv_alloc += nmv_adv*len;
1.103 noro 3812: #endif
1.61 noro 3813: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3814: ndl_copy(DL(t),DL(m));
1.113 noro 3815: CQ(m) = CQ(t);
1.11 noro 3816: }
1.61 noro 3817: MKNDV(NV(p),m0,len,d);
1.14 noro 3818: SG(d) = SG(p);
1.11 noro 3819: return d;
3820: }
3821:
1.61 noro 3822: ND ndvtond(int mod,NDV p)
1.11 noro 3823: {
1.61 noro 3824: ND d;
3825: NM m,m0;
1.11 noro 3826: NMV t;
1.61 noro 3827: int i,len;
1.11 noro 3828:
1.34 noro 3829: if ( !p ) return 0;
1.11 noro 3830: m0 = 0;
1.61 noro 3831: len = p->len;
3832: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3833: NEXTNM(m0,m);
3834: ndl_copy(DL(t),DL(m));
1.113 noro 3835: CQ(m) = CQ(t);
1.11 noro 3836: }
3837: NEXT(m) = 0;
1.61 noro 3838: MKND(NV(p),m0,len,d);
1.14 noro 3839: SG(d) = SG(p);
1.11 noro 3840: return d;
3841: }
3842:
1.3 noro 3843: void ndv_print(NDV p)
3844: {
3845: NMV m;
3846: int i,len;
3847:
1.34 noro 3848: if ( !p ) printf("0\n");
1.3 noro 3849: else {
1.14 noro 3850: len = LEN(p);
1.3 noro 3851: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3852: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3853: else printf("+%d*",CM(m));
1.16 noro 3854: ndl_print(DL(m));
3855: }
3856: printf("\n");
3857: }
3858: }
3859:
1.113 noro 3860: void ndv_print_q(NDV p)
1.16 noro 3861: {
3862: NMV m;
3863: int i,len;
3864:
1.34 noro 3865: if ( !p ) printf("0\n");
1.16 noro 3866: else {
3867: len = LEN(p);
3868: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3869: printf("+");
1.113 noro 3870: printexpr(CO,(Obj)CQ(m));
1.16 noro 3871: printf("*");
1.14 noro 3872: ndl_print(DL(m));
1.3 noro 3873: }
3874: printf("\n");
3875: }
1.25 noro 3876: }
3877:
1.61 noro 3878: NODE ndv_reducebase(NODE x)
1.27 noro 3879: {
3880: int len,i,j;
3881: NDV *w;
3882: NODE t,t0;
3883:
3884: len = length(x);
3885: w = (NDV *)ALLOCA(len*sizeof(NDV));
3886: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3887: for ( i = 0; i < len; i++ ) {
3888: for ( j = 0; j < i; j++ ) {
3889: if ( w[i] && w[j] )
3890: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3891: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3892: }
3893: }
3894: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3895: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3896: }
3897: NEXT(t) = 0; x = t0;
3898: return x;
1.11 noro 3899: }
1.32 noro 3900:
1.43 noro 3901: /* XXX incomplete */
3902:
1.32 noro 3903: void nd_init_ord(struct order_spec *ord)
3904: {
1.43 noro 3905: switch ( ord->id ) {
1.32 noro 3906: case 0:
1.43 noro 3907: switch ( ord->ord.simple ) {
3908: case 0:
3909: nd_dcomp = 1;
3910: nd_isrlex = 1;
3911: break;
3912: case 1:
3913: nd_dcomp = 1;
3914: nd_isrlex = 0;
3915: break;
3916: case 2:
3917: nd_dcomp = 0;
3918: nd_isrlex = 0;
1.45 noro 3919: ndl_compare_function = ndl_lex_compare;
1.58 noro 3920: break;
3921: case 11:
3922: /* XXX */
3923: nd_dcomp = 0;
3924: nd_isrlex = 1;
3925: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3926: break;
3927: default:
3928: error("nd_gr : unsupported order");
3929: }
1.32 noro 3930: break;
3931: case 1:
1.96 noro 3932: /* block order */
1.43 noro 3933: /* XXX */
3934: nd_dcomp = -1;
1.32 noro 3935: nd_isrlex = 0;
1.45 noro 3936: ndl_compare_function = ndl_block_compare;
1.34 noro 3937: break;
1.43 noro 3938: case 2:
1.96 noro 3939: /* matrix order */
3940: /* XXX */
3941: nd_dcomp = -1;
3942: nd_isrlex = 0;
3943: nd_matrix_len = ord->ord.matrix.row;
3944: nd_matrix = ord->ord.matrix.matrix;
3945: ndl_compare_function = ndl_matrix_compare;
3946: break;
3947: case 3:
1.97 noro 3948: /* composite order */
3949: nd_dcomp = -1;
3950: nd_isrlex = 0;
3951: nd_worb_len = ord->ord.composite.length;
3952: nd_worb = ord->ord.composite.w_or_b;
3953: ndl_compare_function = ndl_composite_compare;
1.32 noro 3954: break;
3955: }
1.41 noro 3956: nd_ord = ord;
1.32 noro 3957: }
3958:
1.43 noro 3959: BlockMask nd_create_blockmask(struct order_spec *ord)
3960: {
3961: int n,i,j,s,l;
1.61 noro 3962: UINT *t;
1.43 noro 3963: BlockMask bm;
3964:
1.96 noro 3965: /* we only create mask table for block order */
3966: if ( ord->id != 1 )
1.43 noro 3967: return 0;
3968: n = ord->ord.block.length;
3969: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3970: bm->n = n;
3971: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3972: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3973: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3974: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3975: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3976: l = bm->order_pair[i].length;
3977: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3978: }
3979: return bm;
1.57 noro 3980: }
3981:
3982: EPOS nd_create_epos(struct order_spec *ord)
3983: {
3984: int i,j,l,s,ord_l,ord_o;
3985: EPOS epos;
3986: struct order_pair *op;
3987:
3988: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3989: switch ( ord->id ) {
3990: case 0:
3991: if ( nd_isrlex ) {
3992: for ( i = 0; i < nd_nvar; i++ ) {
3993: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3994: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3995: }
3996: } else {
3997: for ( i = 0; i < nd_nvar; i++ ) {
3998: epos[i].i = nd_exporigin + i/nd_epw;
3999: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4000: }
4001: }
4002: break;
4003: case 1:
4004: /* block order */
4005: l = ord->ord.block.length;
4006: op = ord->ord.block.order_pair;
4007: for ( j = 0, s = 0; j < l; j++ ) {
4008: ord_o = op[j].order;
4009: ord_l = op[j].length;
4010: if ( !ord_o )
4011: for ( i = 0; i < ord_l; i++ ) {
4012: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
4013: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
4014: }
4015: else
4016: for ( i = 0; i < ord_l; i++ ) {
4017: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4018: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4019: }
4020: s += ord_l;
4021: }
4022: break;
4023: case 2:
1.96 noro 4024: /* matrix order */
4025: case 3:
4026: /* composite order */
4027: for ( i = 0; i < nd_nvar; i++ ) {
4028: epos[i].i = nd_exporigin + i/nd_epw;
4029: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4030: }
4031: break;
1.57 noro 4032: }
4033: return epos;
1.43 noro 4034: }
1.59 noro 4035:
4036: /* external interface */
4037:
4038: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4039: {
1.61 noro 4040: NODE t,in0,in;
1.59 noro 4041: ND nd,nf;
1.61 noro 4042: NDV ndv;
4043: VL vv,tv;
4044: int stat,nvar,max,e;
1.113 noro 4045: union oNDC dn;
1.59 noro 4046:
1.116 noro 4047: if ( !f ) {
4048: *rp = 0;
4049: return;
4050: }
1.59 noro 4051: pltovl(v,&vv);
1.74 noro 4052: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4053:
4054: /* get the degree bound */
4055: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4056: for ( tv = vv; tv; tv = NEXT(tv) ) {
4057: e = getdeg(tv->v,(P)BDY(t));
4058: max = MAX(e,max);
4059: }
4060: for ( tv = vv; tv; tv = NEXT(tv) ) {
4061: e = getdeg(tv->v,f);
4062: max = MAX(e,max);
4063: }
4064:
1.59 noro 4065: nd_init_ord(ord);
1.61 noro 4066: nd_setup_parameters(nvar,max);
4067:
4068: /* conversion to ndv */
4069: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4070: NEXTNODE(in0,in);
4071: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4072: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4073: }
4074: NEXTNODE(in0,in);
4075: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4076: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4077: NEXT(in) = 0;
4078:
1.128 noro 4079: /* dont sort */
4080: ndv_setup(m,0,in0,1);
1.59 noro 4081: nd_psn--;
4082: nd_scale=2;
4083: while ( 1 ) {
4084: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4085: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4086: if ( !stat ) {
4087: nd_psn++;
1.103 noro 4088: nd_reconstruct(0,0);
1.59 noro 4089: nd_psn--;
4090: } else
4091: break;
4092: }
1.61 noro 4093: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4094: }
4095:
4096: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4097: {
4098: NM m;
4099: UINT *t,*s;
4100: int i;
4101:
4102: for ( i = 0; i < n; i++ ) r[i] = 0;
4103: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4104: t = DL(m);
4105: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4106: r[i] = CM(m);
4107: }
4108: for ( i = 0; !r[i]; i++ );
4109: return i;
4110: }
4111:
1.113 noro 4112: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4113: {
1.107 noro 4114: NM m;
1.74 noro 4115: UINT *t,*s;
1.107 noro 4116: int i;
1.74 noro 4117:
4118: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4119: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4120: t = DL(m);
4121: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4122: r[i] = CQ(m);
1.74 noro 4123: }
4124: for ( i = 0; !r[i]; i++ );
4125: return i;
4126: }
4127:
1.129 noro 4128: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4129: {
4130: NM m;
4131: NMV mr;
4132: UINT *d,*t,*s;
4133: NDV p;
4134: int i,j,len;
4135: Q *r;
4136:
4137: m = pair->mul;
4138: d = DL(m);
4139: p = nd_ps[pair->index];
4140: len = LEN(p);
4141: r = (Q *)CALLOC(n,sizeof(Q));
4142: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4143: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4144: ndl_add(d,DL(mr),t);
4145: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4146: r[i] = CQ(mr);
4147: }
4148: return r;
4149: }
4150:
1.67 noro 4151: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4152: {
4153: NM m;
4154: NMV mr;
4155: UINT *d,*t,*s;
4156: NDV p;
1.67 noro 4157: unsigned char *ivc;
4158: unsigned short *ivs;
1.81 noro 4159: UINT *v,*ivi,*s0v;
1.67 noro 4160: int i,j,len,prev,diff,cdiff;
4161: IndArray r;
1.64 noro 4162:
4163: m = pair->mul;
4164: d = DL(m);
4165: p = nd_ps[pair->index];
4166: len = LEN(p);
4167: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4168: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4169: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4170: ndl_add(d,DL(mr),t);
4171: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4172: v[j] = i;
4173: }
1.81 noro 4174: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4175: r->head = v[0];
4176: diff = 0;
4177: for ( i = 1; i < len; i++ ) {
4178: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4179: }
4180: if ( diff < 256 ) {
4181: r->width = 1;
4182: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4183: r->index.c = ivc;
4184: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4185: } else if ( diff < 65536 ) {
4186: r->width = 2;
4187: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4188: r->index.s = ivs;
4189: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4190: } else {
4191: r->width = 4;
4192: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4193: r->index.i = ivi;
4194: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4195: }
1.67 noro 4196: return r;
1.64 noro 4197: }
4198:
1.135 noro 4199: int compress_array(Q *svect,Q *cvect,int n)
4200: {
4201: int i,j;
4202:
4203: for ( i = j = 0; i < n; i++ )
4204: if ( svect[i] ) cvect[j++] = svect[i];
4205: return j;
4206: }
4207:
4208: void expand_array(Q *svect,Q *cvect,int n)
4209: {
4210: int i,j;
4211:
4212: for ( i = j = 0; j < n; i++ )
4213: if ( svect[i] ) svect[i] = cvect[j++];
4214: }
4215:
1.133 noro 4216: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4217: {
1.135 noro 4218: int i,j,k,len,pos,prev,nz;
1.113 noro 4219: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4220: IndArray ivect;
4221: unsigned char *ivc;
4222: unsigned short *ivs;
4223: unsigned int *ivi;
4224: NDV redv;
4225: NMV mr;
4226: NODE rp;
4227: int maxrs;
1.133 noro 4228: double hmag;
1.135 noro 4229: Q *cvect;
1.107 noro 4230:
4231: maxrs = 0;
1.134 noro 4232: for ( i = 0; i < col && !svect[i]; i++ );
4233: if ( i == col ) return maxrs;
4234: hmag = p_mag((P)svect[i])*nd_scale;
1.135 noro 4235: cvect = (Q *)ALLOCA(col*sizeof(Q));
1.107 noro 4236: for ( i = 0; i < nred; i++ ) {
4237: ivect = imat[i];
4238: k = ivect->head;
4239: if ( svect[k] ) {
4240: maxrs = MAX(maxrs,rp0[i]->sugar);
1.133 noro 4241: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
1.107 noro 4242: len = LEN(redv); mr = BDY(redv);
1.113 noro 4243: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4244: chsgnq(cs,&mcs);
4245: if ( !UNIQ(cr) ) {
4246: for ( j = 0; j < col; j++ ) {
4247: mulq(svect[j],cr,&c1); svect[j] = c1;
4248: }
4249: }
1.107 noro 4250: svect[k] = 0; prev = k;
4251: switch ( ivect->width ) {
4252: case 1:
4253: ivc = ivect->index.c;
4254: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4255: pos = prev+ivc[j]; prev = pos;
1.113 noro 4256: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4257: }
4258: break;
4259: case 2:
4260: ivs = ivect->index.s;
4261: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4262: pos = prev+ivs[j]; prev = pos;
1.113 noro 4263: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4264: }
4265: break;
4266: case 4:
4267: ivi = ivect->index.i;
4268: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4269: pos = prev+ivi[j]; prev = pos;
1.113 noro 4270: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4271: }
4272: break;
4273: }
1.134 noro 4274: for ( j = k+1; j < col && !svect[j]; j++ );
4275: if ( j == col ) break;
4276: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
1.135 noro 4277: nz = compress_array(svect,cvect,col);
4278: removecont_array(cvect,nz);
4279: expand_array(svect,cvect,nz);
1.134 noro 4280: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
4281: }
1.107 noro 4282: }
1.133 noro 4283: }
1.135 noro 4284: nz = compress_array(svect,cvect,col);
4285: removecont_array(cvect,nz);
4286: expand_array(svect,cvect,nz);
1.133 noro 4287: if ( DP_Print ) {
4288: fprintf(asir_out,"-"); fflush(asir_out);
1.107 noro 4289: }
4290: return maxrs;
4291: }
4292:
1.76 noro 4293: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4294: {
1.67 noro 4295: int i,j,k,len,pos,prev;
1.66 noro 4296: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4297: IndArray ivect;
4298: unsigned char *ivc;
4299: unsigned short *ivs;
4300: unsigned int *ivi;
1.65 noro 4301: NDV redv;
1.67 noro 4302: NMV mr;
1.65 noro 4303: NODE rp;
1.76 noro 4304: int maxrs;
1.65 noro 4305:
1.76 noro 4306: maxrs = 0;
1.74 noro 4307: for ( i = 0; i < nred; i++ ) {
1.65 noro 4308: ivect = imat[i];
1.67 noro 4309: k = ivect->head; svect[k] %= m;
1.65 noro 4310: if ( c = svect[k] ) {
1.76 noro 4311: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4312: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4313: len = LEN(redv); mr = BDY(redv);
4314: svect[k] = 0; prev = k;
4315: switch ( ivect->width ) {
4316: case 1:
4317: ivc = ivect->index.c;
4318: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4319: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4320: prev = pos;
4321: DMA(c1,c,c2,up,lo);
4322: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4323: } else svect[pos] = lo;
4324: }
4325: break;
4326: case 2:
4327: ivs = ivect->index.s;
4328: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4329: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4330: prev = pos;
4331: DMA(c1,c,c2,up,lo);
4332: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4333: } else svect[pos] = lo;
4334: }
4335: break;
4336: case 4:
4337: ivi = ivect->index.i;
4338: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4339: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4340: prev = pos;
4341: DMA(c1,c,c2,up,lo);
4342: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4343: } else svect[pos] = lo;
4344: }
4345: break;
1.65 noro 4346: }
4347: }
4348: }
1.66 noro 4349: for ( i = 0; i < col; i++ )
4350: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4351: return maxrs;
1.65 noro 4352: }
4353:
1.76 noro 4354: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4355: {
4356: int i,j,k,len,pos,prev;
4357: UINT c,c1,c2,c3,up,lo,dmy;
4358: IndArray ivect;
4359: unsigned char *ivc;
4360: unsigned short *ivs;
4361: unsigned int *ivi;
4362: NDV redv;
4363: NMV mr;
4364: NODE rp;
1.76 noro 4365: int maxrs;
1.72 noro 4366:
1.76 noro 4367: maxrs = 0;
1.74 noro 4368: for ( i = 0; i < nred; i++ ) {
1.72 noro 4369: ivect = imat[i];
4370: k = ivect->head; svect[k] %= m;
4371: if ( c = svect[k] ) {
1.76 noro 4372: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4373: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4374: len = LEN(redv); mr = BDY(redv);
4375: svect[k] = 0; prev = k;
4376: switch ( ivect->width ) {
4377: case 1:
4378: ivc = ivect->index.c;
4379: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4380: pos = prev+ivc[j]; prev = pos;
4381: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4382: }
4383: break;
4384: case 2:
4385: ivs = ivect->index.s;
4386: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4387: pos = prev+ivs[j]; prev = pos;
4388: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4389: }
4390: break;
4391: case 4:
4392: ivi = ivect->index.i;
4393: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4394: pos = prev+ivi[j]; prev = pos;
4395: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4396: }
4397: break;
4398: }
4399: }
4400: }
1.76 noro 4401: return maxrs;
1.72 noro 4402: }
4403:
1.65 noro 4404: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4405: {
4406: int j,k,len;
4407: UINT *p;
4408: UINT c;
4409: NDV r;
4410: NMV mr0,mr;
4411:
4412: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4413: if ( !len ) return 0;
4414: else {
1.74 noro 4415: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4416: #if 0
1.74 noro 4417: ndv_alloc += nmv_adv*len;
1.103 noro 4418: #endif
1.65 noro 4419: mr = mr0;
4420: p = s0vect;
4421: for ( j = k = 0; j < col; j++, p += nd_wpd )
4422: if ( !rhead[j] ) {
4423: if ( c = vect[k++] ) {
4424: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4425: }
4426: }
4427: MKNDV(nd_nvar,mr0,len,r);
4428: return r;
4429: }
4430: }
4431:
1.129 noro 4432: /* for preprocessed vector */
4433:
1.113 noro 4434: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4435: {
4436: int j,k,len;
4437: UINT *p;
1.113 noro 4438: Q c;
1.107 noro 4439: NDV r;
4440: NMV mr0,mr;
4441:
4442: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4443: if ( !len ) return 0;
4444: else {
1.109 noro 4445: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4446: #if 0
4447: ndv_alloc += nmv_adv*len;
4448: #endif
4449: mr = mr0;
4450: p = s0vect;
4451: for ( j = k = 0; j < col; j++, p += nd_wpd )
4452: if ( !rhead[j] ) {
4453: if ( c = vect[k++] ) {
1.113 noro 4454: if ( DN(c) )
4455: error("afo");
4456: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4457: }
4458: }
4459: MKNDV(nd_nvar,mr0,len,r);
4460: return r;
4461: }
4462: }
4463:
1.129 noro 4464: /* for plain vector */
4465:
4466: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4467: {
4468: int j,k,len;
4469: UINT *p;
4470: Q c;
4471: NDV r;
4472: NMV mr0,mr;
4473:
4474: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4475: if ( !len ) return 0;
4476: else {
4477: mr0 = (NMV)GC_malloc(nmv_adv*len);
4478: #if 0
4479: ndv_alloc += nmv_adv*len;
4480: #endif
4481: mr = mr0;
4482: p = s0vect;
4483: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4484: if ( c = vect[k] ) {
4485: if ( DN(c) )
4486: error("afo");
4487: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4488: }
4489: MKNDV(nd_nvar,mr0,len,r);
4490: return r;
4491: }
4492: }
4493:
1.133 noro 4494: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 4495: {
4496: ND_pairs t;
4497: NODE sp0,sp;
4498: int stat;
4499: ND spol;
4500:
4501: for ( t = l; t; t = NEXT(t) ) {
1.133 noro 4502: stat = nd_sp(m,trace,t,&spol);
1.65 noro 4503: if ( !stat ) return 0;
4504: if ( spol ) {
4505: add_pbucket_symbolic(bucket,spol);
4506: }
4507: }
1.68 noro 4508: return 1;
1.65 noro 4509: }
4510:
1.133 noro 4511: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 4512: {
4513: NODE rp0,rp;
4514: NM mul,head,s0,s;
1.76 noro 4515: int index,col,i,sugar;
1.65 noro 4516: RHist h;
4517: UINT *s0v,*p;
4518: NM_ind_pair pair;
4519: ND red;
1.133 noro 4520: NDV *ps;
1.65 noro 4521:
4522: s0 = 0; rp0 = 0; col = 0;
1.133 noro 4523: ps = trace?nd_ps_trace:nd_ps;
1.65 noro 4524: while ( 1 ) {
4525: head = remove_head_pbucket_symbolic(bucket);
4526: if ( !head ) break;
4527: if ( !s0 ) s0 = head;
4528: else NEXT(s) = head;
4529: s = head;
4530: index = ndl_find_reducer(DL(head));
4531: if ( index >= 0 ) {
4532: h = nd_psh[index];
4533: NEWNM(mul);
4534: ndl_sub(DL(head),DL(h),DL(mul));
4535: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.133 noro 4536: sugar = TD(DL(mul))+SG(ps[index]);
1.76 noro 4537: MKNM_ind_pair(pair,mul,index,sugar);
1.133 noro 4538: red = ndv_mul_nm_symbolic(mul,ps[index]);
1.65 noro 4539: add_pbucket_symbolic(bucket,nd_remove_head(red));
4540: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4541: }
4542: col++;
4543: }
1.72 noro 4544: if ( rp0 ) NEXT(rp) = 0;
4545: NEXT(s) = 0;
1.65 noro 4546: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4547: for ( i = 0, p = s0v, s = s0; i < col;
4548: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4549: *s0vect = s0v;
4550: *r = rp0;
4551: return col;
4552: }
4553:
1.69 noro 4554: NODE nd_f4(int m)
4555: {
4556: int i,nh,stat,index;
4557: NODE r,g;
4558: ND_pairs d,l,t;
4559: ND spol,red;
4560: NDV nf,redv;
4561: NM s0,s;
1.84 noro 4562: NODE rp0,srp0,nflist;
1.69 noro 4563: int nsp,nred,col,rank,len,k,j,a;
4564: UINT c;
1.74 noro 4565: UINT **spmat;
1.69 noro 4566: UINT *s0vect,*svect,*p,*v;
4567: int *colstat;
4568: IndArray *imat;
4569: int *rhead;
4570: int spcol,sprow;
4571: int sugar;
4572: PGeoBucket bucket;
4573: struct oEGT eg0,eg1,eg_f4;
4574:
1.103 noro 4575: #if 0
1.74 noro 4576: ndv_alloc = 0;
1.103 noro 4577: #endif
1.69 noro 4578: g = 0; d = 0;
4579: for ( i = 0; i < nd_psn; i++ ) {
4580: d = update_pairs(d,g,i);
4581: g = update_base(g,i);
4582: }
4583: while ( d ) {
4584: get_eg(&eg0);
4585: l = nd_minsugarp(d,&d);
4586: sugar = SG(l);
4587: bucket = create_pbucket();
1.133 noro 4588: stat = nd_sp_f4(m,0,l,bucket);
1.69 noro 4589: if ( !stat ) {
4590: for ( t = l; NEXT(t); t = NEXT(t) );
4591: NEXT(t) = d; d = l;
1.103 noro 4592: d = nd_reconstruct(0,d);
1.69 noro 4593: continue;
4594: }
1.86 noro 4595: if ( bucket->m < 0 ) continue;
1.133 noro 4596: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
1.69 noro 4597: if ( !col ) {
4598: for ( t = l; NEXT(t); t = NEXT(t) );
4599: NEXT(t) = d; d = l;
1.103 noro 4600: d = nd_reconstruct(0,d);
1.69 noro 4601: continue;
4602: }
4603: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4604: if ( DP_Print )
4605: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4606: sugar,eg_f4.exectime+eg_f4.gctime);
4607: if ( 1 )
1.133 noro 4608: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
1.74 noro 4609: else
1.133 noro 4610: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
1.69 noro 4611: /* adding new bases */
1.74 noro 4612: for ( r = nflist; r; r = NEXT(r) ) {
4613: nf = (NDV)BDY(r);
1.69 noro 4614: ndv_removecont(m,nf);
1.129 noro 4615: if ( !m && nd_nalg ) {
4616: ND nf1;
4617:
4618: nf1 = ndvtond(m,nf);
4619: nd_monic(0,&nf1);
4620: nd_removecont(m,nf1);
4621: nf = ndtondv(m,nf1);
4622: }
1.77 noro 4623: nh = ndv_newps(m,nf,0);
1.69 noro 4624: d = update_pairs(d,g,nh);
4625: g = update_base(g,nh);
4626: }
4627: }
4628: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4629: #if 0
1.74 noro 4630: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4631: #endif
1.69 noro 4632: return g;
4633: }
1.74 noro 4634:
1.133 noro 4635: NODE nd_f4_trace(int m)
4636: {
4637: int i,nh,stat,index;
4638: NODE r,g;
4639: ND_pairs d,l,l0,t;
4640: ND spol,red;
4641: NDV nf,redv,nfqv,nfv;
4642: NM s0,s;
4643: NODE rp0,srp0,nflist;
4644: int nsp,nred,col,rank,len,k,j,a;
4645: UINT c;
4646: UINT **spmat;
4647: UINT *s0vect,*svect,*p,*v;
4648: int *colstat;
4649: IndArray *imat;
4650: int *rhead;
4651: int spcol,sprow;
4652: int sugar;
4653: PGeoBucket bucket;
4654: struct oEGT eg0,eg1,eg_f4;
4655:
4656: g = 0; d = 0;
4657: for ( i = 0; i < nd_psn; i++ ) {
4658: d = update_pairs(d,g,i);
4659: g = update_base(g,i);
4660: }
4661: while ( d ) {
4662: get_eg(&eg0);
4663: l = nd_minsugarp(d,&d);
4664: sugar = SG(l);
4665: bucket = create_pbucket();
4666: stat = nd_sp_f4(m,0,l,bucket);
4667: if ( !stat ) {
4668: for ( t = l; NEXT(t); t = NEXT(t) );
4669: NEXT(t) = d; d = l;
4670: d = nd_reconstruct(1,d);
4671: continue;
4672: }
4673: if ( bucket->m < 0 ) continue;
4674: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
4675: if ( !col ) {
4676: for ( t = l; NEXT(t); t = NEXT(t) );
4677: NEXT(t) = d; d = l;
4678: d = nd_reconstruct(1,d);
4679: continue;
4680: }
4681: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4682: if ( DP_Print )
4683: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4684: sugar,eg_f4.exectime+eg_f4.gctime);
4685: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
4686: if ( !l0 ) continue;
4687: l = l0;
4688:
4689: /* over Q */
4690: bucket = create_pbucket();
4691: stat = nd_sp_f4(0,1,l,bucket);
4692: if ( !stat ) {
4693: for ( t = l; NEXT(t); t = NEXT(t) );
4694: NEXT(t) = d; d = l;
4695: d = nd_reconstruct(1,d);
4696: continue;
4697: }
4698: if ( bucket->m < 0 ) continue;
4699: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
4700: if ( !col ) {
4701: for ( t = l; NEXT(t); t = NEXT(t) );
4702: NEXT(t) = d; d = l;
4703: d = nd_reconstruct(1,d);
4704: continue;
4705: }
4706: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
4707: /* adding new bases */
4708: for ( r = nflist; r; r = NEXT(r) ) {
4709: nfqv = (NDV)BDY(r);
4710: ndv_removecont(0,nfqv);
4711: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
4712: if ( nd_nalg ) {
4713: ND nf1;
4714:
4715: nf1 = ndvtond(m,nfqv);
4716: nd_monic(0,&nf1);
4717: nd_removecont(0,nf1);
4718: nfqv = ndtondv(0,nf1); nd_free(nf1);
4719: }
4720: nfv = ndv_dup(0,nfqv);
4721: ndv_mod(m,nfv);
4722: ndv_removecont(m,nfv);
4723: nh = ndv_newps(0,nfv,nfqv);
4724: d = update_pairs(d,g,nh);
4725: g = update_base(g,nh);
4726: }
4727: }
4728: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
4729: #if 0
4730: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
4731: #endif
4732: return g;
4733: }
4734:
4735: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 4736: {
1.67 noro 4737: IndArray *imat;
1.106 noro 4738: int nsp,nred,i;
1.65 noro 4739: int *rhead;
1.106 noro 4740: NODE r0,rp;
1.74 noro 4741: ND_pairs sp;
4742: NM_ind_pair *rvect;
1.63 noro 4743:
1.74 noro 4744: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4745: nred = length(rp0);
1.74 noro 4746: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4747: rhead = (int *)ALLOCA(col*sizeof(int));
4748: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4749:
1.74 noro 4750: /* construction of index arrays */
4751: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4752: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4753: rvect[i] = (NM_ind_pair)BDY(rp);
4754: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4755: rhead[imat[i]->head] = 1;
4756: }
1.107 noro 4757: if ( m )
1.133 noro 4758: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
1.107 noro 4759: else
1.133 noro 4760: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4761: return r0;
4762: }
1.74 noro 4763:
1.106 noro 4764: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 4765: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 4766: {
4767: int spcol,sprow,a;
4768: int i,j,k,l,rank;
4769: NODE r0,r;
4770: ND_pairs sp;
4771: ND spol;
4772: int **spmat;
4773: UINT *svect,*v;
4774: int *colstat;
4775: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4776: int maxrs;
4777: int *spsugar;
1.133 noro 4778: ND_pairs *spactive;
1.106 noro 4779:
4780: spcol = col-nred;
4781: get_eg(&eg0);
1.74 noro 4782: /* elimination (1st step) */
4783: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4784: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4785: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.133 noro 4786: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
1.74 noro 4787: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4788: nd_sp(m,0,sp,&spol);
1.75 noro 4789: if ( !spol ) continue;
1.74 noro 4790: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4791: if ( m == -1 )
4792: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4793: else
4794: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4795: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4796: if ( i < col ) {
4797: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4798: for ( j = k = 0; j < col; j++ )
4799: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4800: spsugar[sprow] = MAX(maxrs,SG(spol));
1.133 noro 4801: if ( nz )
4802: spactive[sprow] = sp;
1.74 noro 4803: sprow++;
4804: }
1.76 noro 4805: nd_free(spol);
1.74 noro 4806: }
1.80 noro 4807: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4808: if ( DP_Print ) {
4809: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4810: fflush(asir_out);
4811: }
1.74 noro 4812: /* free index arrays */
4813: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4814:
4815: /* elimination (2nd step) */
4816: colstat = (int *)ALLOCA(spcol*sizeof(int));
4817: if ( m == -1 )
1.76 noro 4818: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4819: else
1.133 noro 4820: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
1.74 noro 4821: r0 = 0;
4822: for ( i = 0; i < rank; i++ ) {
4823: NEXTNODE(r0,r); BDY(r) =
4824: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4825: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4826: GC_free(spmat[i]);
4827: }
1.109 noro 4828: if ( r0 ) NEXT(r) = 0;
1.133 noro 4829:
1.74 noro 4830: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4831: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4832: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4833: if ( DP_Print ) {
1.80 noro 4834: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4835: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4836: nsp,nred,sprow,spcol,rank);
4837: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4838: }
1.133 noro 4839: if ( nz ) {
4840: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
4841: if ( rank > 0 ) {
4842: NEXT(spactive[rank-1]) = 0;
4843: *nz = spactive[0];
4844: } else
4845: *nz = 0;
4846: }
1.74 noro 4847: return r0;
4848: }
4849:
1.133 noro 4850: #if 1
4851: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 4852: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4853: {
4854: int spcol,sprow,a;
4855: int i,j,k,l,rank;
4856: NODE r0,r;
4857: ND_pairs sp;
4858: ND spol;
1.113 noro 4859: Q **spmat;
4860: Q *svect,*v;
1.107 noro 4861: int *colstat;
4862: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4863: int maxrs;
4864: int *spsugar;
1.137 noro 4865: pointer *w;
1.107 noro 4866:
4867: spcol = col-nred;
4868: get_eg(&eg0);
4869: /* elimination (1st step) */
1.113 noro 4870: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4871: svect = (Q *)ALLOCA(col*sizeof(Q));
4872: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4873: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
1.133 noro 4874: nd_sp(0,trace,sp,&spol);
1.107 noro 4875: if ( !spol ) continue;
1.113 noro 4876: nd_to_vect_q(s0vect,col,spol,svect);
1.133 noro 4877: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
1.107 noro 4878: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4879: if ( i < col ) {
1.113 noro 4880: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4881: for ( j = k = 0; j < col; j++ )
4882: if ( !rhead[j] ) v[k++] = svect[j];
4883: spsugar[sprow] = MAX(maxrs,SG(spol));
4884: sprow++;
4885: }
1.109 noro 4886: /* nd_free(spol); */
1.107 noro 4887: }
4888: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4889: if ( DP_Print ) {
4890: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4891: fflush(asir_out);
4892: }
4893: /* free index arrays */
1.109 noro 4894: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4895:
4896: /* elimination (2nd step) */
4897: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4898: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.137 noro 4899: w = (pointer *)ALLOCA(rank*sizeof(pointer));
4900: for ( i = 0; i < rank; i++ ) {
4901: w[rank-i-1] = (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
4902: SG((NDV)w[rank-i-1]) = spsugar[i];
4903: /* GC_free(spmat[i]); */
4904: }
1.138 noro 4905: #if 0
1.137 noro 4906: qsort(w,rank,sizeof(NDV),
4907: (int (*)(const void *,const void *))ndv_compare);
4908: #endif
1.107 noro 4909: r0 = 0;
4910: for ( i = 0; i < rank; i++ ) {
1.137 noro 4911: NEXTNODE(r0,r); BDY(r) = w[i];
1.107 noro 4912: }
1.137 noro 4913: if ( r0 ) NEXT(r) = 0;
1.135 noro 4914:
1.109 noro 4915: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4916: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4917: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4918: if ( DP_Print ) {
4919: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4920: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4921: nsp,nred,sprow,spcol,rank);
4922: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4923: }
4924: return r0;
4925: }
1.129 noro 4926: #else
4927: void printm(Q **mat,int row,int col)
4928: {
4929: int i,j;
4930: printf("[");
4931: for ( i = 0; i < row; i++ ) {
4932: for ( j = 0; j < col; j++ ) {
4933: printexpr(CO,mat[i][j]); printf(" ");
4934: }
4935: printf("]\n");
4936: }
4937: }
4938:
4939: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
4940: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4941: {
4942: int row,a;
4943: int i,j,rank;
4944: NODE r0,r;
4945: ND_pairs sp;
4946: ND spol;
4947: Q **mat;
4948: int *colstat;
4949: int *sugar;
4950:
4951: row = nsp+nred;
4952: /* make the matrix */
4953: mat = (Q **)ALLOCA(row*sizeof(Q *));
4954: sugar = (int *)ALLOCA(row*sizeof(int));
4955: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4956: nd_sp(0,0,sp,&spol);
4957: if ( !spol ) continue;
4958: mat[row] = (Q *)MALLOC(col*sizeof(Q));
4959: nd_to_vect_q(s0vect,col,spol,mat[row]);
4960: sugar[row] = SG(spol);
4961: row++;
4962: }
4963: for ( i = 0; i < nred; i++, row++ ) {
4964: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
4965: sugar[row] = rvect[i]->sugar;
4966: }
4967: /* elimination */
4968: colstat = (int *)ALLOCA(col*sizeof(int));
4969: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
4970: r0 = 0;
4971: for ( i = 0; i < rank; i++ ) {
4972: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
4973: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
4974: if ( rhead[j] ) continue;
4975: NEXTNODE(r0,r); BDY(r) =
4976: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
4977: SG((NDV)BDY(r)) = sugar[i];
4978: }
4979: if ( r0 ) NEXT(r) = 0;
4980: printf("\n");
4981: return r0;
4982: }
4983: #endif
1.107 noro 4984:
1.74 noro 4985: FILE *nd_write,*nd_read;
4986:
4987: void nd_send_int(int a) {
4988: write_int(nd_write,&a);
4989: }
4990:
4991: void nd_send_intarray(int *p,int len) {
4992: write_intarray(nd_write,p,len);
4993: }
4994:
4995: int nd_recv_int() {
4996: int a;
4997:
4998: read_int(nd_read,&a);
4999: return a;
5000: }
5001:
5002: void nd_recv_intarray(int *p,int len) {
5003: read_intarray(nd_read,p,len);
5004: }
5005:
5006: void nd_send_ndv(NDV p) {
5007: int len,i;
5008: NMV m;
5009:
5010: if ( !p ) nd_send_int(0);
5011: else {
5012: len = LEN(p);
5013: nd_send_int(len);
5014: m = BDY(p);
5015: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5016: nd_send_int(CM(m));
5017: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 5018: }
1.74 noro 5019: }
5020: }
5021:
5022: void nd_send_nd(ND p) {
5023: int len,i;
5024: NM m;
5025:
5026: if ( !p ) nd_send_int(0);
5027: else {
5028: len = LEN(p);
5029: nd_send_int(len);
5030: m = BDY(p);
5031: for ( i = 0; i < len; i++, m = NEXT(m) ) {
5032: nd_send_int(CM(m));
5033: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 5034: }
1.74 noro 5035: }
5036: }
1.65 noro 5037:
1.74 noro 5038: NDV nd_recv_ndv()
5039: {
5040: int len,i;
5041: NMV m,m0;
5042: NDV r;
1.65 noro 5043:
1.74 noro 5044: len = nd_recv_int();
5045: if ( !len ) return 0;
5046: else {
5047: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5048: #if 0
1.74 noro 5049: ndv_alloc += len*nmv_adv;
1.103 noro 5050: #endif
1.74 noro 5051: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5052: CM(m) = nd_recv_int();
5053: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 5054: }
1.74 noro 5055: MKNDV(nd_nvar,m0,len,r);
5056: return r;
5057: }
5058: }
1.65 noro 5059:
1.74 noro 5060: int ox_exec_f4_red(Q proc)
5061: {
5062: Obj obj;
5063: STRING fname;
5064: NODE arg;
5065: int s;
5066: extern int ox_need_conv,ox_file_io;
5067:
5068: MKSTR(fname,"nd_exec_f4_red");
5069: arg = mknode(2,proc,fname);
5070: Pox_cmo_rpc(arg,&obj);
5071: s = get_ox_server_id(QTOS(proc));
5072: nd_write = iofp[s].out;
5073: nd_read = iofp[s].in;
5074: ox_need_conv = ox_file_io = 0;
5075: return s;
5076: }
5077:
1.133 noro 5078: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5079: {
5080: int nsp,nred;
5081: int i,rank,s;
5082: NODE rp,r0,r;
5083: ND_pairs sp;
5084: NM_ind_pair pair;
5085: NMV nmv;
5086: NM nm;
5087: NDV nf;
5088: Obj proc,dmy;
5089:
5090: ox_launch_main(0,0,&proc);
5091: s = ox_exec_f4_red((Q)proc);
5092:
5093: nd_send_int(m);
5094: nd_send_int(nd_nvar);
5095: nd_send_int(nd_bpe);
5096: nd_send_int(nd_wpd);
5097: nd_send_int(nmv_adv);
5098:
1.89 noro 5099: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 5100:
5101: nd_send_int(nd_psn);
5102: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5103:
5104: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5105: nd_send_int(nsp);
5106: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5107: nd_send_int(sp->i1); nd_send_int(sp->i2);
5108: }
5109:
5110: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5111:
5112: nred = length(rp0); nd_send_int(nred);
5113: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5114: pair = (NM_ind_pair)BDY(rp);
5115: nd_send_int(pair->index);
5116: nd_send_intarray(pair->mul->dl,nd_wpd);
5117: }
5118: fflush(nd_write);
5119: rank = nd_recv_int();
5120: fprintf(asir_out,"rank=%d\n",rank);
5121: r0 = 0;
5122: for ( i = 0; i < rank; i++ ) {
5123: nf = nd_recv_ndv();
5124: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5125: }
5126: Pox_shutdown(mknode(1,proc),&dmy);
5127: return r0;
5128: }
5129:
5130: /* server side */
5131:
5132: void nd_exec_f4_red_dist()
5133: {
5134: int m,i,nsp,col,s0size,nred,spcol,j,k;
5135: NM_ind_pair *rp0;
5136: NDV nf;
5137: UINT *s0vect;
5138: IndArray *imat;
5139: int *rhead;
5140: int **spmat;
5141: UINT *svect,*v;
5142: ND_pairs *sp0;
5143: int *colstat;
5144: int a,sprow,rank;
1.89 noro 5145: struct order_spec *ord;
1.74 noro 5146: Obj ordspec;
5147: ND spol;
1.76 noro 5148: int maxrs;
5149: int *spsugar;
1.74 noro 5150:
5151: nd_read = iofp[0].in;
5152: nd_write = iofp[0].out;
5153: m = nd_recv_int();
5154: nd_nvar = nd_recv_int();
5155: nd_bpe = nd_recv_int();
5156: nd_wpd = nd_recv_int();
5157: nmv_adv = nd_recv_int();
5158:
5159: loadobj(nd_read,&ordspec);
1.89 noro 5160: create_order_spec(0,ordspec,&ord);
5161: nd_init_ord(ord);
1.74 noro 5162: nd_setup_parameters(nd_nvar,0);
5163:
5164: nd_psn = nd_recv_int();
5165: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5166: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5167: for ( i = 0; i < nd_psn; i++ ) {
5168: nd_ps[i] = nd_recv_ndv();
5169: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5170: }
5171:
5172: nsp = nd_recv_int();
5173: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5174: for ( i = 0; i < nsp; i++ ) {
5175: NEWND_pairs(sp0[i]);
5176: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5177: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5178: }
5179:
5180: col = nd_recv_int();
5181: s0size = col*nd_wpd;
5182: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5183: nd_recv_intarray(s0vect,s0size);
5184:
5185: nred = nd_recv_int();
5186: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5187: for ( i = 0; i < nred; i++ ) {
5188: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5189: rp0[i]->index = nd_recv_int();
5190: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5191: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5192: }
5193:
5194: spcol = col-nred;
5195: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5196: rhead = (int *)MALLOC(col*sizeof(int));
5197: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5198:
5199: /* construction of index arrays */
5200: for ( i = 0; i < nred; i++ ) {
5201: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5202: rhead[imat[i]->head] = 1;
5203: }
5204:
5205: /* elimination (1st step) */
5206: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5207: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5208: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5209: for ( a = sprow = 0; a < nsp; a++ ) {
5210: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5211: if ( !spol ) continue;
1.74 noro 5212: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5213: if ( m == -1 )
5214: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5215: else
5216: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5217: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5218: if ( i < col ) {
5219: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5220: for ( j = k = 0; j < col; j++ )
5221: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5222: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5223: sprow++;
5224: }
1.76 noro 5225: nd_free(spol);
1.74 noro 5226: }
5227: /* elimination (2nd step) */
5228: colstat = (int *)ALLOCA(spcol*sizeof(int));
5229: if ( m == -1 )
1.76 noro 5230: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5231: else
1.133 noro 5232: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
1.74 noro 5233: nd_send_int(rank);
5234: for ( i = 0; i < rank; i++ ) {
5235: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5236: nd_send_ndv(nf);
1.63 noro 5237: }
1.74 noro 5238: fflush(nd_write);
1.107 noro 5239: }
5240:
1.113 noro 5241: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5242: {
1.109 noro 5243: int mod,i,j,t,c,rank,rank0,inv;
5244: int *ci,*ri;
1.113 noro 5245: Q dn;
1.109 noro 5246: MAT m,nm;
1.108 noro 5247: int **wmat;
5248:
5249: /* XXX */
5250: mod = 99999989;
5251: wmat = (int **)ALLOCA(row*sizeof(int *));
5252: for ( i = 0; i < row; i++ ) {
5253: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5254: for ( j = 0; j < col; j++ ) {
1.113 noro 5255: if ( mat0[i][j] ) {
5256: t = rem(NM(mat0[i][j]),mod);
5257: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5258: wmat[i][j] = t;
5259: } else
1.108 noro 5260: wmat[i][j] = 0;
5261: }
5262: }
1.133 noro 5263: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
1.109 noro 5264: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5265: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5266: if ( rank != rank0 )
5267: error("afo");
5268: for ( i = 0; i < row; i++ )
5269: for ( j = 0; j < col; j++ )
5270: mat0[i][j] = 0;
5271: c = col-rank;
5272: for ( i = 0; i < rank; i++ ) {
1.113 noro 5273: mat0[i][ri[i]] = dn;
1.109 noro 5274: for ( j = 0; j < c; j++ )
1.113 noro 5275: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5276: }
1.113 noro 5277: inv = invm(rem(NM(dn),mod),mod);
5278: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5279: for ( i = 0; i < row; i++ )
5280: for ( j = 0; j < col; j++ ) {
1.113 noro 5281: if ( mat0[i][j] ) {
5282: t = rem(NM(mat0[i][j]),mod);
5283: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5284: } else
1.109 noro 5285: t = 0;
5286: c = dmar(t,inv,0,mod);
5287: if ( wmat[i][j] != c )
5288: error("afo");
5289: }
5290: return rank;
1.76 noro 5291: }
5292:
1.133 noro 5293: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5294: {
5295: int i,j,k,l,inv,a,rank,s;
5296: unsigned int *t,*pivot,*pk;
5297: unsigned int **mat;
1.133 noro 5298: ND_pairs pair;
1.76 noro 5299:
5300: mat = (unsigned int **)mat0;
5301: for ( rank = 0, j = 0; j < col; j++ ) {
5302: for ( i = rank; i < row; i++ )
5303: mat[i][j] %= md;
5304: for ( i = rank; i < row; i++ )
5305: if ( mat[i][j] )
5306: break;
5307: if ( i == row ) {
5308: colstat[j] = 0;
5309: continue;
5310: } else
5311: colstat[j] = 1;
5312: if ( i != rank ) {
5313: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5314: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
1.133 noro 5315: if ( spactive ) {
5316: pair = spactive[i]; spactive[i] = spactive[rank];
5317: spactive[rank] = pair;
5318: }
1.76 noro 5319: }
5320: pivot = mat[rank];
5321: s = sugar[rank];
5322: inv = invm(pivot[j],md);
5323: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5324: if ( *pk ) {
5325: if ( *pk >= (unsigned int)md )
5326: *pk %= md;
5327: DMAR(*pk,inv,0,md,*pk)
5328: }
5329: for ( i = rank+1; i < row; i++ ) {
5330: t = mat[i];
5331: if ( a = t[j] ) {
5332: sugar[i] = MAX(sugar[i],s);
5333: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5334: }
5335: }
5336: rank++;
5337: }
5338: for ( j = col-1, l = rank-1; j >= 0; j-- )
5339: if ( colstat[j] ) {
5340: pivot = mat[l];
5341: s = sugar[l];
5342: for ( i = 0; i < l; i++ ) {
5343: t = mat[i];
5344: t[j] %= md;
5345: if ( a = t[j] ) {
5346: sugar[i] = MAX(sugar[i],s);
5347: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5348: }
5349: }
5350: l--;
5351: }
5352: for ( j = 0, l = 0; l < rank; j++ )
5353: if ( colstat[j] ) {
5354: t = mat[l];
5355: for ( k = j; k < col; k++ )
5356: if ( t[k] >= (unsigned int)md )
5357: t[k] %= md;
5358: l++;
5359: }
5360: return rank;
5361: }
5362:
5363: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5364: {
5365: int i,j,k,l,inv,a,rank,s;
5366: unsigned int *t,*pivot,*pk;
5367: unsigned int **mat;
5368:
5369: mat = (unsigned int **)mat0;
5370: for ( rank = 0, j = 0; j < col; j++ ) {
5371: for ( i = rank; i < row; i++ )
5372: if ( mat[i][j] )
5373: break;
5374: if ( i == row ) {
5375: colstat[j] = 0;
5376: continue;
5377: } else
5378: colstat[j] = 1;
5379: if ( i != rank ) {
5380: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5381: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5382: }
5383: pivot = mat[rank];
5384: s = sugar[rank];
5385: inv = _invsf(pivot[j]);
5386: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5387: if ( *pk )
5388: *pk = _mulsf(*pk,inv);
5389: for ( i = rank+1; i < row; i++ ) {
5390: t = mat[i];
5391: if ( a = t[j] ) {
5392: sugar[i] = MAX(sugar[i],s);
5393: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5394: }
5395: }
5396: rank++;
5397: }
5398: for ( j = col-1, l = rank-1; j >= 0; j-- )
5399: if ( colstat[j] ) {
5400: pivot = mat[l];
5401: s = sugar[l];
5402: for ( i = 0; i < l; i++ ) {
5403: t = mat[i];
5404: if ( a = t[j] ) {
5405: sugar[i] = MAX(sugar[i],s);
5406: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5407: }
5408: }
5409: l--;
5410: }
5411: return rank;
1.77 noro 5412: }
5413:
5414: int ndv_ishomo(NDV p)
5415: {
5416: NMV m;
5417: int len,h;
5418:
5419: if ( !p ) return 1;
5420: len = LEN(p);
5421: m = BDY(p);
5422: h = TD(DL(m));
5423: NMV_ADV(m);
5424: for ( len--; len; len--, NMV_ADV(m) )
5425: if ( TD(DL(m)) != h ) return 0;
5426: return 1;
5427: }
5428:
5429: void ndv_save(NDV p,int index)
5430: {
5431: FILE *s;
5432: char name[BUFSIZ];
5433: short id;
5434: int nv,sugar,len,n,i,td,e,j;
5435: NMV m;
5436: unsigned int *dl;
5437:
5438: sprintf(name,"%s/%d",Demand,index);
5439: s = fopen(name,"w");
5440: savevl(s,0);
5441: if ( !p ) {
5442: saveobj(s,0);
5443: return;
5444: }
5445: id = O_DP;
5446: nv = NV(p);
5447: sugar = SG(p);
5448: len = LEN(p);
5449: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5450: write_int(s,&len);
5451:
5452: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5453: saveobj(s,(Obj)CQ(m));
1.77 noro 5454: dl = DL(m);
5455: td = TD(dl);
5456: write_int(s,&td);
5457: for ( j = 0; j < nv; j++ ) {
5458: e = GET_EXP(dl,j);
5459: write_int(s,&e);
5460: }
5461: }
5462: fclose(s);
5463: }
5464:
5465: NDV ndv_load(int index)
5466: {
5467: FILE *s;
5468: char name[BUFSIZ];
5469: short id;
5470: int nv,sugar,len,n,i,td,e,j;
5471: NDV d;
5472: NMV m0,m;
5473: unsigned int *dl;
5474: Obj obj;
5475:
5476: sprintf(name,"%s/%d",Demand,index);
5477: s = fopen(name,"r");
5478: if ( !s ) return 0;
5479:
5480: skipvl(s);
5481: read_short(s,&id);
5482: if ( !id ) return 0;
5483: read_int(s,&nv);
5484: read_int(s,&sugar);
5485: read_int(s,&len);
5486:
5487: m0 = m = MALLOC(len*nmv_adv);
5488: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5489: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5490: dl = DL(m);
5491: ndl_zero(dl);
5492: read_int(s,&td); TD(dl) = td;
5493: for ( j = 0; j < nv; j++ ) {
5494: read_int(s,&e);
5495: PUT_EXP(dl,j,e);
5496: }
5497: if ( nd_blockmask ) ndl_weight_mask(dl);
5498: }
5499: fclose(s);
5500: MKNDV(nv,m0,len,d);
5501: SG(d) = sugar;
5502: return d;
1.99 noro 5503: }
5504:
1.102 noro 5505: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5506: {
5507: VL fv,tv;
1.102 noro 5508: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5509: pointer **m;
1.113 noro 5510: Q mone;
1.125 noro 5511: P **w;
1.126 noro 5512: P mp,r;
1.102 noro 5513: NDV **dm;
5514: NDV *t,*mi,*mj;
5515: NDV d,s,mij,mjj;
5516: ND u;
5517: NMV nmv;
1.114 noro 5518: UINT *bound;
1.99 noro 5519: PGeoBucket bucket;
5520: struct order_spec *ord;
1.126 noro 5521: Q dq,dt,ds;
5522: N gn,qn,dn0,nm,dn;
1.99 noro 5523:
5524: create_order_spec(0,0,&ord);
5525: nd_init_ord(ord);
5526: get_vars((Obj)f,&fv);
5527: if ( f->row != f->col )
5528: error("nd_det : non-square matrix");
5529: n = f->row;
1.125 noro 5530: m = f->body;
1.99 noro 5531: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5532:
5533: if ( !nvar ) {
5534: if ( !mod )
5535: detp(CO,(P **)m,n,rp);
5536: else {
5537: w = (P **)almat_pointer(n,n);
5538: for ( i = 0; i < n; i++ )
5539: for ( j = 0; j < n; j++ )
5540: ptomp(mod,(P)m[i][j],&w[i][j]);
5541: detmp(CO,mod,w,n,&mp);
5542: mptop(mp,rp);
5543: }
5544: return;
5545: }
1.126 noro 5546:
5547: if ( !mod ) {
5548: w = (P **)almat_pointer(n,n);
5549: dq = ONE;
5550: for ( i = 0; i < n; i++ ) {
5551: dn0 = ONEN;
5552: for ( j = 0; j < n; j++ ) {
5553: if ( !m[i][j] ) continue;
5554: lgp(m[i][j],&nm,&dn);
5555: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5556: }
5557: if ( !UNIN(dn0) ) {
5558: NTOQ(dn0,1,ds);
5559: for ( j = 0; j < n; j++ )
5560: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5561: mulq(dq,ds,&dt); dq = dt;
5562: } else
5563: for ( j = 0; j < n; j++ )
5564: w[i][j] = (P)m[i][j];
5565: }
5566: m = (pointer **)w;
5567: }
1.125 noro 5568:
1.99 noro 5569: for ( i = 0, max = 0; i < n; i++ )
5570: for ( j = 0; j < n; j++ )
5571: for ( tv = fv; tv; tv = NEXT(tv) ) {
5572: e = getdeg(tv->v,(P)m[i][j]);
5573: max = MAX(e,max);
5574: }
1.114 noro 5575: nd_setup_parameters(nvar,max);
1.102 noro 5576: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5577: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5578: for ( j = 0; j < n; j++ ) {
5579: dm[i][j] = ptondv(CO,fv,m[i][j]);
5580: if ( mod ) ndv_mod(mod,dm[i][j]);
5581: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5582: }
5583: d = ptondv(CO,fv,(P)ONE);
5584: if ( mod ) ndv_mod(mod,d);
1.113 noro 5585: chsgnq(ONE,&mone);
1.99 noro 5586: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5587: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5588: for ( i = j; i < n && !dm[i][j]; i++ );
5589: if ( i == n ) {
5590: *rp = 0;
5591: return;
5592: }
1.102 noro 5593: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5594: for ( k = j; k < n; k++ )
5595: for ( l = j; l < n; l++ )
1.102 noro 5596: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5597: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5598: }
5599: if ( k0 != j ) {
5600: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5601: sgn = -sgn;
5602: }
5603: if ( l0 != j ) {
5604: for ( k = j; k < n; k++ ) {
5605: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5606: }
5607: sgn = -sgn;
5608: }
1.114 noro 5609: bound = nd_det_compute_bound(dm,n,j);
5610: if ( ndl_check_bound(bound,bound) )
5611: nd_det_reconstruct(dm,n,j,d);
5612:
1.99 noro 5613: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5614: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5615: mi = dm[i]; mij = mi[j];
1.102 noro 5616: if ( mod )
5617: ndv_mul_c(mod,mij,mod-1);
5618: else
1.113 noro 5619: ndv_mul_c_q(mij,mone);
1.99 noro 5620: for ( k = j+1; k < n; k++ ) {
1.114 noro 5621: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5622: bucket = create_pbucket();
1.104 noro 5623: if ( mi[k] ) {
1.102 noro 5624: nmv = BDY(mjj); len = LEN(mjj);
5625: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5626: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5627: add_pbucket(mod,bucket,u);
1.99 noro 5628: }
1.104 noro 5629: }
1.99 noro 5630: if ( mj[k] && mij ) {
1.102 noro 5631: nmv = BDY(mij); len = LEN(mij);
5632: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5633: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5634: add_pbucket(mod,bucket,u);
1.99 noro 5635: }
5636: }
1.104 noro 5637: u = nd_quo(mod,bucket,d);
1.102 noro 5638: mi[k] = ndtondv(mod,u);
1.99 noro 5639: }
1.114 noro 5640: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5641: }
5642: d = mjj;
5643: }
1.125 noro 5644: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5645: if ( sgn < 0 )
1.102 noro 5646: if ( mod )
5647: ndv_mul_c(mod,d,mod-1);
5648: else
1.113 noro 5649: ndv_mul_c_q(d,mone);
1.126 noro 5650: r = ndvtop(mod,CO,fv,d);
5651: if ( !mod && !UNIQ(dq) )
5652: divsp(CO,r,(P)dq,rp);
5653: else
5654: *rp = r;
1.99 noro 5655: }
5656:
1.102 noro 5657: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5658: {
5659: NM mr,mr0;
1.102 noro 5660: NM tnm;
5661: NMV m;
1.99 noro 5662: UINT *d0,*dt,*dm;
5663: int c,n,td,i,c1,c2,len;
1.113 noro 5664: Q q;
1.99 noro 5665: ND r;
5666:
5667: if ( !p ) return 0;
5668: else {
1.102 noro 5669: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5670: d0 = DL(m0);
1.102 noro 5671: td = TD(d);
1.99 noro 5672: mr0 = 0;
5673: NEWNM(tnm);
1.102 noro 5674: if ( mod ) {
5675: c = CM(m0);
5676: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5677: ndl_add(DL(m),d0,DL(tnm));
5678: if ( ndl_reducible(DL(tnm),d) ) {
5679: NEXTNM(mr0,mr);
5680: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5681: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5682: }
5683: }
5684: } else {
1.113 noro 5685: q = CQ(m0);
1.102 noro 5686: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5687: ndl_add(DL(m),d0,DL(tnm));
5688: if ( ndl_reducible(DL(tnm),d) ) {
5689: NEXTNM(mr0,mr);
1.113 noro 5690: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5691: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5692: }
1.99 noro 5693: }
5694: }
5695: if ( !mr0 )
5696: return 0;
5697: else {
5698: NEXT(mr) = 0;
1.105 noro 5699: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5700: MKND(NV(p),mr0,len,r);
5701: SG(r) = SG(p) + TD(d0);
5702: return r;
5703: }
5704: }
1.114 noro 5705: }
5706:
5707: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5708: {
5709: int i,obpe,oadv,h,k,l;
5710: static NM prev_nm_free_list;
5711: EPOS oepos;
5712:
5713: obpe = nd_bpe;
5714: oadv = nmv_adv;
5715: oepos = nd_epos;
5716: if ( obpe < 2 ) nd_bpe = 2;
5717: else if ( obpe < 3 ) nd_bpe = 3;
5718: else if ( obpe < 4 ) nd_bpe = 4;
5719: else if ( obpe < 5 ) nd_bpe = 5;
5720: else if ( obpe < 6 ) nd_bpe = 6;
5721: else if ( obpe < 8 ) nd_bpe = 8;
5722: else if ( obpe < 10 ) nd_bpe = 10;
5723: else if ( obpe < 16 ) nd_bpe = 16;
5724: else if ( obpe < 32 ) nd_bpe = 32;
5725: else error("nd_det_reconstruct : exponent too large");
5726:
5727: nd_setup_parameters(nd_nvar,0);
5728: prev_nm_free_list = _nm_free_list;
5729: _nm_free_list = 0;
5730: for ( k = j; k < n; k++ )
5731: for (l = j; l < n; l++ )
5732: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5733: ndv_realloc(d,obpe,oadv,oepos);
5734: prev_nm_free_list = 0;
5735: #if 0
5736: GC_gcollect();
5737: #endif
5738: }
5739:
5740: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5741: {
5742: UINT *d0,*d1,*d,*t,*r;
5743: int k,l;
5744:
5745: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5746: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5747: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5748: for ( k = j; k < n; k++ )
5749: for ( l = j; l < n; l++ )
5750: if ( dm[k][l] ) {
5751: d = ndv_compute_bound(dm[k][l]);
5752: ndl_lcm(d,d0,d1);
5753: t = d1; d1 = d0; d0 = t;
5754: }
5755: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5756: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5757: return r;
1.117 noro 5758: }
5759:
5760: DL nd_separate_d(UINT *d,UINT *trans)
5761: {
1.121 noro 5762: int n,td,i,e,j;
1.117 noro 5763: DL a;
5764:
5765: ndl_zero(trans);
5766: td = 0;
1.121 noro 5767: for ( i = 0; i < nd_ntrans; i++ ) {
5768: e = GET_EXP(d,i);
5769: PUT_EXP(trans,i,e);
5770: td += MUL_WEIGHT(e,i);
5771: }
5772: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5773: /* homogenized */
5774: i = nd_nvar-1;
1.117 noro 5775: e = GET_EXP(d,i);
5776: PUT_EXP(trans,i,e);
5777: td += MUL_WEIGHT(e,i);
5778: }
5779: TD(trans) = td;
5780: if ( nd_blockmask) ndl_weight_mask(trans);
5781: NEWDL(a,nd_nalg);
5782: td = 0;
1.121 noro 5783: for ( i = 0; i < nd_nalg; i++ ) {
5784: j = nd_ntrans+i;
5785: e = GET_EXP(d,j);
5786: a->d[i] = e;
1.117 noro 5787: td += e;
5788: }
5789: a->td = td;
5790: return a;
5791: }
5792:
1.118 noro 5793: int nd_monic(int mod,ND *p)
1.117 noro 5794: {
5795: UINT *trans,*t;
5796: DL alg;
5797: MP mp0,mp;
1.118 noro 5798: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5799: ND r;
1.117 noro 5800: DL dl;
5801: DP nm;
5802: NDV ndv;
1.118 noro 5803: DAlg inv,cd;
1.117 noro 5804: ND s,c;
1.118 noro 5805: Q l,mul;
5806: N ln;
5807: int n,ntrans,i,e,td,is_lc,len;
5808: NumberField nf;
5809: struct oEGT eg0,eg1;
5810:
5811: if ( !(nf = get_numberfield()) )
5812: error("nd_monic : current_numberfield is not set");
1.117 noro 5813:
1.118 noro 5814: /* Q coef -> DAlg coef */
5815: NEWNM(ma0); ma = ma0;
5816: m = BDY(*p);
5817: is_lc = 1;
5818: while ( 1 ) {
5819: NEWMP(mp0); mp = mp0;
5820: mp->c = (P)CQ(m);
5821: mp->dl = nd_separate_d(DL(m),DL(ma));
5822: NEWNM(mb);
5823: for ( m = NEXT(m); m; m = NEXT(m) ) {
5824: alg = nd_separate_d(DL(m),DL(mb));
5825: if ( !ndl_equal(DL(ma),DL(mb)) )
5826: break;
5827: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5828: }
5829: NEXT(mp) = 0;
5830: MKDP(nd_nalg,mp0,nm);
5831: MKDAlg(nm,ONE,cd);
5832: if ( is_lc == 1 ) {
5833: /* if the lc is a rational number, we have nothing to do */
5834: if ( !mp0->dl->td )
5835: return 1;
5836:
5837: get_eg(&eg0);
5838: invdalg(cd,&inv);
5839: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5840: /* check the validity of inv */
5841: if ( mod && !rem(NM(inv->dn),mod) )
5842: return 0;
5843: CA(ma) = nf->one;
5844: is_lc = 0;
5845: ln = ONEN;
5846: } else {
5847: muldalg(cd,inv,&CA(ma));
5848: lcmn(ln,NM(CA(ma)->dn),&ln);
5849: }
5850: if ( m ) {
5851: NEXT(ma) = mb; ma = mb;
5852: } else {
5853: NEXT(ma) = 0;
5854: break;
5855: }
5856: }
5857: /* l = lcm(denoms) */
5858: NTOQ(ln,1,l);
5859: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5860: divq(l,CA(m)->dn,&mul);
5861: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5862: NEXTNM(mr0,mr);
5863: mulq((Q)mp->c,mul,&CQ(mr));
5864: dl = mp->dl;
5865: td = TD(DL(m));
5866: ndl_copy(DL(m),DL(mr));
1.121 noro 5867: for ( i = 0; i < nd_nalg; i++ ) {
5868: e = dl->d[i];
5869: PUT_EXP(DL(mr),i+nd_ntrans,e);
5870: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5871: }
5872: TD(DL(mr)) = td;
5873: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5874: }
1.117 noro 5875: }
1.118 noro 5876: NEXT(mr) = 0;
5877: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5878: MKND(NV(*p),mr0,len,r);
5879: /* XXX */
5880: SG(r) = SG(*p);
5881: nd_free(*p);
5882: *p = r;
5883: return 1;
1.59 noro 5884: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>