Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.142
1.142 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.141 2006/06/17 10:12:06 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:
1.142 ! noro 1551: for ( i = nd_psn-1; i >= 0; i-- ) {
! 1552: if ( SG(nd_psh[i]) != sugar ) continue;
1.92 noro 1553: if ( nd_demand )
1554: nfv = ndv_load(i);
1555: else
1556: nfv = nd_ps[i];
1557: s = ndvtond(m,nfv);
1558: s = nd_separate_head(s,&head);
1.122 noro 1559: stat = nd_nf(m,s,nd_ps,1,&dn,&nf);
1560: if ( !stat ) return 0;
1.92 noro 1561: if ( !m ) {
1.113 noro 1562: NTOQ(NM(dn.z),SGN(dn.z),num);
1563: mulq(HCQ(head),num,&q); HCQ(head) = q;
1564: if ( DN(dn.z) ) {
1565: NTOQ(DN(dn.z),1,den);
1566: nd_mul_c_q(nf,den);
1567: }
1.92 noro 1568: }
1569: nf = nd_add(m,head,nf);
1570: ndv_free(nfv);
1571: nd_removecont(m,nf);
1572: nfv = ndtondv(m,nf);
1573: nd_free(nf);
1574: nd_bound[i] = ndv_compute_bound(nfv);
1575: if ( nd_demand ) {
1576: ndv_save(nfv,i);
1577: ndv_free(nfv);
1578: } else
1579: nd_ps[i] = nfv;
1580: }
1.122 noro 1581: return 1;
1.92 noro 1582: }
1583:
1.27 noro 1584: /* return value = 0 => input is not a GB */
1585:
1.77 noro 1586: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1587: {
1588: int i,nh,sugar,stat;
1.23 noro 1589: NODE r,g,t;
1.1 noro 1590: ND_pairs d;
1591: ND_pairs l;
1.117 noro 1592: ND h,nf,s,head,nf1;
1.63 noro 1593: NDV nfv;
1.113 noro 1594: Q q,num,den;
1595: union oNDC dn;
1.140 noro 1596: int diag_count = 0;
1.1 noro 1597:
1.23 noro 1598: g = 0; d = 0;
1599: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1600: d = update_pairs(d,g,i);
1601: g = update_base(g,i);
1602: }
1603: sugar = 0;
1604: while ( d ) {
1605: again:
1606: l = nd_minp(d,&d);
1.14 noro 1607: if ( SG(l) != sugar ) {
1.122 noro 1608: if ( ishomo ) {
1.140 noro 1609: diag_count = 0;
1.122 noro 1610: stat = do_diagonalize(sugar,m);
1611: if ( !stat ) {
1612: NEXT(l) = d; d = l;
1613: d = nd_reconstruct(0,d);
1614: goto again;
1615: }
1616: }
1.14 noro 1617: sugar = SG(l);
1.71 noro 1618: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1619: }
1.53 noro 1620: stat = nd_sp(m,0,l,&h);
1.1 noro 1621: if ( !stat ) {
1622: NEXT(l) = d; d = l;
1.103 noro 1623: d = nd_reconstruct(0,d);
1.1 noro 1624: goto again;
1625: }
1.41 noro 1626: #if USE_GEOBUCKET
1.69 noro 1627: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1628: #else
1.69 noro 1629: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1630: #endif
1.1 noro 1631: if ( !stat ) {
1632: NEXT(l) = d; d = l;
1.103 noro 1633: d = nd_reconstruct(0,d);
1.1 noro 1634: goto again;
1635: } else if ( nf ) {
1.27 noro 1636: if ( checkonly ) return 0;
1.71 noro 1637: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1638: nd_removecont(m,nf);
1.118 noro 1639: if ( !m && nd_nalg ) {
1640: nd_monic(0,&nf);
1641: nd_removecont(m,nf);
1.117 noro 1642: }
1.63 noro 1643: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1644: nh = ndv_newps(m,nfv,0);
1.141 noro 1645: if ( !m && (ishomo && ++diag_count == diag_period) ) {
1.140 noro 1646: diag_count = 0;
1647: stat = do_diagonalize(sugar,m);
1648: if ( !stat ) {
1649: NEXT(l) = d; d = l;
1650: d = nd_reconstruct(1,d);
1651: goto again;
1652: }
1653: }
1.1 noro 1654: d = update_pairs(d,g,nh);
1655: g = update_base(g,nh);
1656: FREENDP(l);
1657: } else {
1.71 noro 1658: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1659: FREENDP(l);
1660: }
1661: }
1.77 noro 1662: if ( nd_demand )
1663: for ( t = g; t; t = NEXT(t) )
1664: BDY(t) = (pointer)ndv_load((int)BDY(t));
1665: else
1666: for ( t = g; t; t = NEXT(t) )
1667: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1668: return g;
1669: }
1670:
1.122 noro 1671: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1672: {
1673: int i,nh,stat;
1674: NODE r,g,t;
1675: ND h,nf,nfq,s,head;
1676: NDV nfv,nfqv;
1.113 noro 1677: Q q,den,num;
1678: union oNDC dn;
1.91 noro 1679:
1.142 ! noro 1680: for ( i = nd_psn-1; i >= 0; i-- ) {
! 1681: if ( SG(nd_psh[i]) != sugar ) continue;
1.91 noro 1682: /* for nd_ps */
1683: s = ndvtond(m,nd_ps[i]);
1684: s = nd_separate_head(s,&head);
1.122 noro 1685: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1686: if ( !stat ) return 0;
1.91 noro 1687: nf = nd_add(m,head,nf);
1688: ndv_free(nd_ps[i]);
1689: nd_ps[i] = ndtondv(m,nf);
1690: nd_free(nf);
1691:
1692: /* for nd_ps_trace */
1693: if ( nd_demand )
1694: nfv = ndv_load(i);
1695: else
1696: nfv = nd_ps_trace[i];
1697: s = ndvtond(0,nfv);
1698: s = nd_separate_head(s,&head);
1.122 noro 1699: stat = nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1700: if ( !stat ) return 0;
1.113 noro 1701: NTOQ(NM(dn.z),SGN(dn.z),num);
1702: mulq(HCQ(head),num,&q); HCQ(head) = q;
1703: if ( DN(dn.z) ) {
1704: NTOQ(DN(dn.z),1,den);
1705: nd_mul_c_q(nf,den);
1706: }
1.91 noro 1707: nf = nd_add(0,head,nf);
1708: ndv_free(nfv);
1709: nd_removecont(0,nf);
1710: nfv = ndtondv(0,nf);
1711: nd_free(nf);
1712: nd_bound[i] = ndv_compute_bound(nfv);
1713: if ( nd_demand ) {
1714: ndv_save(nfv,i);
1715: ndv_free(nfv);
1716: } else
1717: nd_ps_trace[i] = nfv;
1718: }
1.122 noro 1719: return 1;
1.91 noro 1720: }
1721:
1.118 noro 1722: static struct oEGT eg_invdalg;
1723: struct oEGT eg_le;
1724:
1.82 noro 1725: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1726: {
1727: int i,nh,sugar,stat;
1.23 noro 1728: NODE r,g,t;
1.20 noro 1729: ND_pairs d;
1730: ND_pairs l;
1.82 noro 1731: ND h,nf,nfq,s,head;
1.63 noro 1732: NDV nfv,nfqv;
1.113 noro 1733: Q q,den,num;
1734: union oNDC dn;
1.118 noro 1735: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1736: int diag_count = 0;
1.20 noro 1737:
1.118 noro 1738: init_eg(&eg_monic);
1739: init_eg(&eg_invdalg);
1740: init_eg(&eg_le);
1.23 noro 1741: g = 0; d = 0;
1742: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1743: d = update_pairs(d,g,i);
1744: g = update_base(g,i);
1745: }
1746: sugar = 0;
1747: while ( d ) {
1748: again:
1749: l = nd_minp(d,&d);
1750: if ( SG(l) != sugar ) {
1.130 noro 1751: #if 1
1.122 noro 1752: if ( ishomo ) {
1753: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1754: diag_count = 0;
1.122 noro 1755: if ( !stat ) {
1756: NEXT(l) = d; d = l;
1757: d = nd_reconstruct(1,d);
1758: goto again;
1759: }
1760: }
1.130 noro 1761: #endif
1.20 noro 1762: sugar = SG(l);
1.71 noro 1763: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1764: }
1.53 noro 1765: stat = nd_sp(m,0,l,&h);
1.20 noro 1766: if ( !stat ) {
1767: NEXT(l) = d; d = l;
1.103 noro 1768: d = nd_reconstruct(1,d);
1.20 noro 1769: goto again;
1770: }
1.41 noro 1771: #if USE_GEOBUCKET
1.53 noro 1772: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1773: #else
1.69 noro 1774: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1775: #endif
1.20 noro 1776: if ( !stat ) {
1777: NEXT(l) = d; d = l;
1.103 noro 1778: d = nd_reconstruct(1,d);
1.20 noro 1779: goto again;
1780: } else if ( nf ) {
1.77 noro 1781: if ( nd_demand ) {
1782: nfqv = ndv_load(nd_psn);
1783: nfq = ndvtond(0,nfqv);
1784: } else
1785: nfq = 0;
1786: if ( !nfq ) {
1787: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1788: NEXT(l) = d; d = l;
1.103 noro 1789: d = nd_reconstruct(1,d);
1.77 noro 1790: goto again;
1791: }
1.70 noro 1792: }
1.20 noro 1793: if ( nfq ) {
1.70 noro 1794: /* m|HC(nfq) => failure */
1.113 noro 1795: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1.63 noro 1796:
1.71 noro 1797: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1798: if ( nd_nalg ) {
1799: /* m|DN(HC(nf)^(-1)) => failure */
1800: get_eg(&egm0);
1801: if ( !nd_monic(m,&nfq) ) return 0;
1802: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1803: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1804: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1805: } else {
1806: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1807: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1808: }
1.77 noro 1809: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1810: if ( ishomo && ++diag_count == diag_period ) {
1811: diag_count = 0;
1812: stat = do_diagonalize_trace(sugar,m);
1813: if ( !stat ) {
1814: NEXT(l) = d; d = l;
1815: d = nd_reconstruct(1,d);
1816: goto again;
1817: }
1818: }
1.20 noro 1819: d = update_pairs(d,g,nh);
1820: g = update_base(g,nh);
1821: } else {
1.71 noro 1822: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1823: }
1824: } else {
1.71 noro 1825: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1826: }
1827: FREENDP(l);
1828: }
1.77 noro 1829: if ( nd_demand )
1830: for ( t = g; t; t = NEXT(t) )
1831: BDY(t) = (pointer)ndv_load((int)BDY(t));
1832: else
1833: for ( t = g; t; t = NEXT(t) )
1834: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1835: if ( nd_nalg ) {
1836: print_eg("monic",&eg_monic);
1837: print_eg("invdalg",&eg_invdalg);
1838: print_eg("le",&eg_le);
1839: }
1.20 noro 1840: return g;
1841: }
1842:
1.23 noro 1843: int ndv_compare(NDV *p1,NDV *p2)
1844: {
1.34 noro 1845: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1846: }
1847:
1848: int ndv_compare_rev(NDV *p1,NDV *p2)
1849: {
1.34 noro 1850: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1851: }
1852:
1.61 noro 1853: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1854: {
1.69 noro 1855: int i,n,stat;
1856: ND nf,g,head;
1.23 noro 1857: NODE t,a0,a;
1.113 noro 1858: union oNDC dn;
1.69 noro 1859: NDV *w;
1.113 noro 1860: Q q,num,den;
1.23 noro 1861:
1.69 noro 1862: n = length(f);
1863: #if 0
1864: w = (NDV *)ALLOCA(n*sizeof(NDV));
1865: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1866: qsort(w,n,sizeof(NDV),
1867: (int (*)(const void *,const void *))ndv_compare);
1868: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1869: #endif
1.128 noro 1870: ndv_setup(m,0,f,0);
1.69 noro 1871: for ( i = 0; i < n; ) {
1872: g = ndvtond(m,nd_ps[i]);
1873: g = nd_separate_head(g,&head);
1874: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1875: if ( !stat )
1.103 noro 1876: nd_reconstruct(0,0);
1.69 noro 1877: else {
1.71 noro 1878: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1879: if ( !m ) {
1.113 noro 1880: NTOQ(NM(dn.z),SGN(dn.z),num);
1881: mulq(HCQ(head),num,&q); HCQ(head) = q;
1882: if ( DN(dn.z) ) {
1883: NTOQ(DN(dn.z),1,den);
1884: nd_mul_c_q(nf,den);
1885: }
1.85 noro 1886: }
1.69 noro 1887: nf = nd_add(m,head,nf);
1888: ndv_free(nd_ps[i]);
1.24 noro 1889: nd_removecont(m,nf);
1.69 noro 1890: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1891: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1892: i++;
1.23 noro 1893: }
1894: }
1.71 noro 1895: if ( DP_Print ) { printf("\n"); }
1.23 noro 1896: for ( a0 = 0, i = 0; i < n; i++ ) {
1897: NEXTNODE(a0,a);
1.69 noro 1898: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1899: }
1900: NEXT(a) = 0;
1901: return a0;
1902: }
1903:
1.1 noro 1904: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1905: {
1906: ND_pairs d1,nd,cur,head,prev,remove;
1907:
1908: if ( !g ) return d;
1909: d = crit_B(d,t);
1910: d1 = nd_newpairs(g,t);
1911: d1 = crit_M(d1);
1912: d1 = crit_F(d1);
1.55 noro 1913: if ( do_weyl )
1914: head = d1;
1915: else {
1916: prev = 0; cur = head = d1;
1917: while ( cur ) {
1918: if ( crit_2( cur->i1,cur->i2 ) ) {
1919: remove = cur;
1920: if ( !prev ) head = cur = NEXT(cur);
1921: else cur = NEXT(prev) = NEXT(cur);
1922: FREENDP(remove);
1923: } else {
1924: prev = cur; cur = NEXT(cur);
1925: }
1.1 noro 1926: }
1927: }
1928: if ( !d )
1929: return head;
1930: else {
1931: nd = d;
1.34 noro 1932: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1933: NEXT(nd) = head;
1934: return d;
1935: }
1936: }
1937:
1938: ND_pairs nd_newpairs( NODE g, int t )
1939: {
1940: NODE h;
1.61 noro 1941: UINT *dl;
1.34 noro 1942: int ts,s;
1.1 noro 1943: ND_pairs r,r0;
1944:
1.20 noro 1945: dl = DL(nd_psh[t]);
1.34 noro 1946: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1947: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1948: NEXTND_pairs(r0,r);
1949: r->i1 = (int)BDY(h);
1950: r->i2 = t;
1.20 noro 1951: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1952: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1953: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1954: }
1955: NEXT(r) = 0;
1956: return r0;
1957: }
1958:
1959: ND_pairs crit_B( ND_pairs d, int s )
1960: {
1961: ND_pairs cur,head,prev,remove;
1.61 noro 1962: UINT *t,*tl,*lcm;
1.1 noro 1963: int td,tdl;
1964:
1965: if ( !d ) return 0;
1.20 noro 1966: t = DL(nd_psh[s]);
1.1 noro 1967: prev = 0;
1968: head = cur = d;
1.61 noro 1969: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1970: while ( cur ) {
1971: tl = cur->lcm;
1972: if ( ndl_reducible(tl,t)
1.20 noro 1973: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1974: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1975: remove = cur;
1976: if ( !prev ) {
1977: head = cur = NEXT(cur);
1978: } else {
1979: cur = NEXT(prev) = NEXT(cur);
1980: }
1981: FREENDP(remove);
1982: } else {
1.34 noro 1983: prev = cur; cur = NEXT(cur);
1.1 noro 1984: }
1985: }
1986: return head;
1987: }
1988:
1989: ND_pairs crit_M( ND_pairs d1 )
1990: {
1991: ND_pairs e,d2,d3,dd,p;
1.61 noro 1992: UINT *id,*jd;
1.1 noro 1993:
1994: for ( dd = 0, e = d1; e; e = d3 ) {
1995: if ( !(d2 = NEXT(e)) ) {
1996: NEXT(e) = dd;
1997: return e;
1998: }
1.34 noro 1999: id = LCM(e);
1.1 noro 2000: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2001: p = NEXT(d2);
2002: jd = LCM(d2);
2003: if ( ndl_equal(jd,id) )
2004: ;
2005: else if ( TD(jd) > TD(id) )
1.1 noro 2006: if ( ndl_reducible(jd,id) ) continue;
2007: else ;
1.34 noro 2008: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2009: NEXT(d2) = d3;
2010: d3 = d2;
2011: }
2012: NEXT(e) = dd;
2013: dd = e;
2014: continue;
2015: /**/
2016: delit: NEXT(d2) = d3;
2017: d3 = d2;
2018: for ( ; p; p = d2 ) {
2019: d2 = NEXT(p);
2020: NEXT(p) = d3;
2021: d3 = p;
2022: }
2023: FREENDP(e);
2024: }
2025: return dd;
2026: }
2027:
2028: ND_pairs crit_F( ND_pairs d1 )
2029: {
2030: ND_pairs rest, head,remove;
2031: ND_pairs last, p, r, w;
2032: int s;
2033:
2034: for ( head = last = 0, p = d1; NEXT(p); ) {
2035: r = w = equivalent_pairs(p,&rest);
1.14 noro 2036: s = SG(r);
1.1 noro 2037: w = NEXT(w);
2038: while ( w ) {
2039: if ( crit_2(w->i1,w->i2) ) {
2040: r = w;
2041: w = NEXT(w);
2042: while ( w ) {
2043: remove = w;
2044: w = NEXT(w);
2045: FREENDP(remove);
2046: }
2047: break;
1.14 noro 2048: } else if ( SG(w) < s ) {
1.1 noro 2049: FREENDP(r);
2050: r = w;
1.14 noro 2051: s = SG(r);
1.1 noro 2052: w = NEXT(w);
2053: } else {
2054: remove = w;
2055: w = NEXT(w);
2056: FREENDP(remove);
2057: }
2058: }
2059: if ( last ) NEXT(last) = r;
2060: else head = r;
2061: NEXT(last = r) = 0;
2062: p = rest;
2063: if ( !p ) return head;
2064: }
2065: if ( !last ) return p;
2066: NEXT(last) = p;
2067: return head;
2068: }
2069:
2070: int crit_2( int dp1, int dp2 )
2071: {
1.20 noro 2072: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2073: }
2074:
1.40 noro 2075: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2076: {
2077: ND_pairs w,p,r,s;
1.61 noro 2078: UINT *d;
1.1 noro 2079:
2080: w = d1;
1.34 noro 2081: d = LCM(w);
1.1 noro 2082: s = NEXT(w);
2083: NEXT(w) = 0;
2084: for ( r = 0; s; s = p ) {
2085: p = NEXT(s);
1.34 noro 2086: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2087: NEXT(s) = w; w = s;
1.1 noro 2088: } else {
1.39 noro 2089: NEXT(s) = r; r = s;
1.1 noro 2090: }
2091: }
2092: *prest = r;
2093: return w;
2094: }
2095:
2096: NODE update_base(NODE nd,int ndp)
2097: {
1.61 noro 2098: UINT *dl, *dln;
1.137 noro 2099: NODE last, p, head;
1.1 noro 2100:
1.20 noro 2101: dl = DL(nd_psh[ndp]);
1.1 noro 2102: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2103: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2104: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2105: p = NEXT(p);
2106: if ( last ) NEXT(last) = p;
2107: } else {
2108: if ( !last ) head = p;
2109: p = NEXT(last = p);
2110: }
2111: }
2112: head = append_one(head,ndp);
2113: return head;
2114: }
2115:
2116: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2117: {
2118: ND_pairs m,ml,p,l;
1.61 noro 2119: UINT *lcm;
1.33 noro 2120: int s,td,len,tlen,c,c1;
1.1 noro 2121:
2122: if ( !(p = NEXT(m = d)) ) {
2123: *prest = p;
2124: NEXT(m) = 0;
2125: return m;
2126: }
1.14 noro 2127: s = SG(m);
1.77 noro 2128: if ( !NoSugar ) {
2129: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2130: if ( (SG(p) < s)
2131: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2132: ml = l; m = p; s = SG(m);
2133: }
2134: } else {
2135: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2136: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2137: ml = l; m = p; s = SG(m);
2138: }
2139: }
1.1 noro 2140: if ( !ml ) *prest = NEXT(m);
2141: else {
2142: NEXT(ml) = NEXT(m);
2143: *prest = d;
2144: }
2145: NEXT(m) = 0;
2146: return m;
2147: }
2148:
1.63 noro 2149: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2150: {
1.69 noro 2151: int msugar,i;
1.63 noro 2152: ND_pairs t,dm0,dm,dr0,dr;
2153:
2154: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2155: if ( SG(t) < msugar ) msugar = SG(t);
2156: dm0 = 0; dr0 = 0;
1.69 noro 2157: for ( i = 0, t = d; t; t = NEXT(t) )
2158: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2159: if ( dm0 ) NEXT(dm) = t;
2160: else dm0 = t;
2161: dm = t;
1.69 noro 2162: i++;
1.63 noro 2163: } else {
2164: if ( dr0 ) NEXT(dr) = t;
2165: else dr0 = t;
2166: dr = t;
2167: }
2168: NEXT(dm) = 0;
2169: if ( dr0 ) NEXT(dr) = 0;
2170: *prest = dr0;
2171: return dm0;
2172: }
2173:
1.77 noro 2174: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2175: {
1.3 noro 2176: int len;
1.13 noro 2177: RHist r;
1.20 noro 2178: NDV b;
1.3 noro 2179:
1.1 noro 2180: if ( nd_psn == nd_pslen ) {
2181: nd_pslen *= 2;
1.11 noro 2182: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2183: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2184: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2185: nd_bound = (UINT **)
2186: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2187: }
1.39 noro 2188: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2189: nd_ps[nd_psn] = a;
1.39 noro 2190: if ( aq ) {
1.63 noro 2191: nd_ps_trace[nd_psn] = aq;
2192: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2193: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2194: } else {
1.63 noro 2195: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2196: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2197: }
1.77 noro 2198: if ( nd_demand ) {
2199: if ( aq ) {
2200: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2201: nd_ps_trace[nd_psn] = 0;
2202: } else {
2203: ndv_save(nd_ps[nd_psn],nd_psn);
2204: nd_ps[nd_psn] = 0;
2205: }
2206: }
1.1 noro 2207: return nd_psn++;
2208: }
2209:
1.128 noro 2210: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2211: {
1.5 noro 2212: int i,j,td,len,max;
1.1 noro 2213: NODE s,s0,f0;
1.61 noro 2214: UINT *d;
1.13 noro 2215: RHist r;
1.69 noro 2216: NDV *w;
1.61 noro 2217: NDV a,am;
1.11 noro 2218:
2219: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2220:
1.69 noro 2221: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2222: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2223: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.139 noro 2224: if ( !dont_sort ) {
2225: /* XXX heuristic */
2226: if ( !nd_ord->id && (nd_ord->ord.simple<2) )
2227: qsort(w,nd_psn,sizeof(NDV),
2228: (int (*)(const void *,const void *))ndv_compare_rev);
2229: else
2230: qsort(w,nd_psn,sizeof(NDV),
2231: (int (*)(const void *,const void *))ndv_compare);
2232: }
1.69 noro 2233: nd_pslen = 2*nd_psn;
1.11 noro 2234: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2235: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2236: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2237: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2238:
1.11 noro 2239: if ( !nd_red )
1.13 noro 2240: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2241: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2242: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2243: if ( trace ) {
1.69 noro 2244: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2245: ndv_removecont(0,a);
2246: am = nd_ps[i] = ndv_dup(mod,a);
2247: ndv_mod(mod,am);
2248: ndv_removecont(mod,am);
2249: } else {
1.69 noro 2250: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2251: ndv_removecont(mod,a);
1.39 noro 2252: }
1.61 noro 2253: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2254: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2255: nd_psh[i] = r;
1.77 noro 2256: if ( nd_demand ) {
2257: if ( trace ) {
2258: ndv_save(nd_ps_trace[i],i);
2259: nd_ps_trace[i] = 0;
2260: } else {
2261: ndv_save(nd_ps[i],i);
2262: nd_ps[i] = 0;
2263: }
2264: }
1.20 noro 2265: }
2266: }
2267:
1.119 noro 2268: struct order_spec *append_block(struct order_spec *spec,
2269: int nv,int nalg,int ord);
2270:
1.121 noro 2271: extern VECT current_dl_weight_vector_obj;
2272: static VECT prev_weight_vector_obj;
2273:
1.120 noro 2274: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2275: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2276: {
1.121 noro 2277: NODE alist,t,s,r0,r,arg;
1.120 noro 2278: VL tv;
2279: P poly;
2280: DP d;
2281: Alg alpha,dp;
2282: DAlg inv,da,hc;
2283: MP m;
1.121 noro 2284: int i,nvar,nalg,n;
1.120 noro 2285: NumberField nf;
2286: LIST f1,f2;
2287: struct order_spec *current_spec;
1.121 noro 2288: VECT obj,obj0;
2289: Obj tmp;
1.120 noro 2290:
2291: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2292: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2293:
2294: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2295: NEXTNODE(alist,t); MKV(tv->v,poly);
2296: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2297: tv->v = tv->v->priv;
2298: }
2299: NEXT(t) = 0;
2300:
2301: /* simplification, makeing polynomials monic */
2302: setfield_dalg(alist);
2303: obj_algtodalg(f,&f1);
2304: for ( t = BDY(f); t; t = NEXT(t) ) {
2305: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2306: hc = (DAlg)BDY(d)->c;
2307: if ( NID(hc) == N_DA ) {
2308: invdalg(hc,&inv);
2309: for ( m = BDY(d); m; m = NEXT(m) ) {
2310: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2311: }
2312: }
2313: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2314: }
2315: obj_dalgtoalg(f1,&f);
2316:
2317: /* append alg vars to the var list */
2318: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2319: NEXT(tv) = av;
2320:
2321: /* append a block to ord */
2322: *ord1p = append_block(ord,nvar,nalg,2);
2323:
2324: /* create generator list */
2325: nf = get_numberfield();
2326: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2327: MKAlg(nf->defpoly[i],dp);
2328: MKNODE(s,dp,t); t = s;
2329: }
2330: MKLIST(f1,t);
2331: *alistp = alist;
2332: algobjtorat(f1,f1p);
1.121 noro 2333:
2334: /* creating a new weight vector */
2335: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2336: n = nvar+nalg+1;
2337: MKVECT(obj,n);
2338: if ( obj0 && obj0->len == nvar )
2339: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2340: else
2341: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2342: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2343: BDY(obj)[n-1] = (pointer)ONE;
2344: arg = mknode(1,obj);
2345: Pdp_set_weight(arg,&tmp);
2346: }
2347:
2348: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2349: {
2350: NODE s,t,u0,u;
2351: P p;
2352: VL tv;
2353: Obj obj,tmp;
2354: NODE arg;
2355:
2356: u0 = 0;
2357: for ( t = r; t; t = NEXT(t) ) {
2358: p = (P)BDY(t);
2359: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2360: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2361: }
2362: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2363: NEXTNODE(u0,u);
2364: BDY(u) = (pointer)p;
2365: }
2366: }
2367: arg = mknode(1,prev_weight_vector_obj);
2368: Pdp_set_weight(arg,&tmp);
2369:
2370: return u0;
1.120 noro 2371: }
2372:
1.63 noro 2373: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2374: {
1.119 noro 2375: VL tv,fv,vv,vc,av;
2376: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2377: int e,max,nvar,i;
1.69 noro 2378: NDV b;
1.119 noro 2379: int ishomo,nalg;
2380: Alg alpha,dp;
2381: P p;
2382: LIST f1,f2;
2383: Obj obj;
2384: NumberField nf;
1.120 noro 2385: struct order_spec *ord1;
1.1 noro 2386:
1.78 noro 2387: if ( !m && Demand ) nd_demand = 1;
2388: else nd_demand = 0;
2389:
1.103 noro 2390: #if 0
1.74 noro 2391: ndv_alloc = 0;
1.103 noro 2392: #endif
1.1 noro 2393: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2394: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2395: switch ( ord->id ) {
2396: case 1:
2397: if ( ord->nv != nvar )
2398: error("nd_{gr,f4} : invalid order specification");
2399: break;
2400: default:
2401: break;
2402: }
1.119 noro 2403: nd_nalg = 0;
1.123 noro 2404: av = 0;
1.119 noro 2405: if ( !m ) {
2406: get_algtree((Obj)f,&av);
2407: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2408: nd_ntrans = nvar;
1.119 noro 2409: nd_nalg = nalg;
2410: /* #i -> t#i */
2411: if ( nalg ) {
1.120 noro 2412: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2413: ord = ord1;
2414: f = f1;
1.119 noro 2415: }
1.120 noro 2416: nvar += nalg;
1.119 noro 2417: }
1.32 noro 2418: nd_init_ord(ord);
1.61 noro 2419: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2420: for ( tv = vv; tv; tv = NEXT(tv) ) {
2421: e = getdeg(tv->v,(P)BDY(t));
2422: max = MAX(e,max);
2423: }
2424: nd_setup_parameters(nvar,max);
1.77 noro 2425: ishomo = 1;
1.1 noro 2426: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2427: b = (pointer)ptondv(CO,vv,(P)BDY(t));
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;
2460: P p;
2461: LIST f1,f2;
2462: Obj obj;
2463: NumberField nf;
2464: struct order_spec *ord1;
2465:
2466: get_vars((Obj)f,&fv); pltovl(v,&vv);
2467: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2468: switch ( ord->id ) {
2469: case 1:
2470: if ( ord->nv != nvar )
2471: error("nd_check : invalid order specification");
2472: break;
2473: default:
2474: break;
2475: }
2476: nd_nalg = 0;
2477: av = 0;
2478: if ( !m ) {
2479: get_algtree((Obj)f,&av);
2480: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2481: nd_ntrans = nvar;
2482: nd_nalg = nalg;
2483: /* #i -> t#i */
2484: if ( nalg ) {
2485: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2486: ord = ord1;
2487: f = f1;
2488: }
2489: nvar += nalg;
2490: }
2491: nd_init_ord(ord);
2492: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2493: for ( tv = vv; tv; tv = NEXT(tv) ) {
2494: e = getdeg(tv->v,(P)BDY(t));
2495: max = MAX(e,max);
2496: }
2497: nd_setup_parameters(nvar,max);
2498: ishomo = 1;
2499: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2500: b = (pointer)ptondv(CO,vv,(P)BDY(t));
2501: if ( ishomo )
2502: ishomo = ishomo && ndv_ishomo(b);
2503: if ( m ) ndv_mod(m,b);
2504: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2505: }
2506: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2507: ndv_setup(m,0,fd0,0);
1.127 noro 2508: for ( x = 0, i = 0; i < nd_psn; i++ )
2509: x = update_base(x,i);
2510: if ( do_check ) {
2511: x = nd_gb(m,ishomo,1);
2512: if ( !x ) {
2513: *rp = 0;
2514: return;
2515: }
2516: } else {
2517: for ( t = x; t; t = NEXT(t) )
2518: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2519: }
2520: x = ndv_reducebase(x);
2521: x = ndv_reduceall(m,x);
2522: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2523: NEXTNODE(r0,r);
2524: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2525: }
2526: if ( r0 ) NEXT(r) = 0;
2527: if ( nalg )
2528: r0 = postprocess_algcoef(av,alist,r0);
2529: MKLIST(*rp,r0);
1.20 noro 2530: }
2531:
1.133 noro 2532: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2533: {
1.119 noro 2534: VL tv,fv,vv,vc,av;
2535: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2536: int m,nocheck,nvar,mindex,e,max;
2537: NDV c;
2538: NMV a;
1.27 noro 2539: P p;
1.61 noro 2540: EPOS oepos;
1.119 noro 2541: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2542: Alg alpha,dp;
2543: P poly;
2544: LIST f1,f2;
2545: Obj obj;
2546: NumberField nf;
1.120 noro 2547: struct order_spec *ord1;
1.20 noro 2548:
2549: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2550: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2551: switch ( ord->id ) {
2552: case 1:
2553: if ( ord->nv != nvar )
2554: error("nd_gr_trace : invalid order specification");
2555: break;
2556: default:
2557: break;
2558: }
1.119 noro 2559:
2560: get_algtree((Obj)f,&av);
2561: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2562: nd_ntrans = nvar;
1.119 noro 2563: nd_nalg = nalg;
2564: /* #i -> t#i */
2565: if ( nalg ) {
1.120 noro 2566: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2567: ord = ord1;
2568: f = f1;
1.119 noro 2569: }
1.120 noro 2570: nvar += nalg;
1.119 noro 2571:
1.52 noro 2572: nocheck = 0;
2573: mindex = 0;
2574:
1.78 noro 2575: if ( Demand ) nd_demand = 1;
2576: else nd_demand = 0;
2577:
1.52 noro 2578: /* setup modulus */
2579: if ( trace < 0 ) {
2580: trace = -trace;
2581: nocheck = 1;
2582: }
2583: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2584: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2585: for ( tv = vv; tv; tv = NEXT(tv) ) {
2586: e = getdeg(tv->v,(P)BDY(t));
2587: max = MAX(e,max);
1.23 noro 2588: }
1.61 noro 2589: nd_init_ord(ord);
2590: nd_setup_parameters(nvar,max);
2591: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2592: ishomo = 1;
1.61 noro 2593: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2594: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2595: if ( ishomo )
2596: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2597: if ( c ) {
2598: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2599: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2600: }
1.61 noro 2601: }
2602: if ( in0 ) NEXT(in) = 0;
2603: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2604: if ( !ishomo && homo ) {
1.122 noro 2605: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2606: c = (NDV)BDY(t); len = LEN(c);
2607: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2608: wmax = MAX(TD(DL(a)),wmax);
2609: }
2610: homogenize_order(ord,nvar,&ord1);
1.89 noro 2611: nd_init_ord(ord1);
1.61 noro 2612: nd_setup_parameters(nvar+1,wmax);
2613: for ( t = fd0; t; t = NEXT(t) )
2614: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2615: }
1.52 noro 2616: while ( 1 ) {
1.77 noro 2617: if ( Demand )
2618: nd_demand = 1;
1.128 noro 2619: ndv_setup(m,1,fd0,0);
1.133 noro 2620: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
1.52 noro 2621: if ( !cand ) {
2622: /* failure */
1.61 noro 2623: if ( trace > 1 ) { *rp = 0; return; }
2624: else m = get_lprime(++mindex);
1.52 noro 2625: continue;
2626: }
1.82 noro 2627: if ( !ishomo && homo ) {
1.27 noro 2628: /* dehomogenization */
1.61 noro 2629: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2630: nd_init_ord(ord);
1.61 noro 2631: nd_setup_parameters(nvar,0);
1.27 noro 2632: }
1.77 noro 2633: nd_demand = 0;
1.61 noro 2634: cand = ndv_reducebase(cand);
2635: cand = ndv_reduceall(0,cand);
1.78 noro 2636: cbpe = nd_bpe;
1.61 noro 2637: if ( nocheck )
2638: break;
2639: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2640: /* success */
2641: break;
2642: else if ( trace > 1 ) {
2643: /* failure */
2644: *rp = 0; return;
1.61 noro 2645: } else {
1.52 noro 2646: /* try the next modulus */
2647: m = get_lprime(++mindex);
1.61 noro 2648: /* reset the parameters */
1.82 noro 2649: if ( !ishomo && homo ) {
1.89 noro 2650: nd_init_ord(ord1);
1.61 noro 2651: nd_setup_parameters(nvar+1,wmax);
2652: } else {
2653: nd_init_ord(ord);
2654: nd_setup_parameters(nvar,max);
2655: }
2656: }
1.52 noro 2657: }
1.27 noro 2658: /* dp->p */
1.61 noro 2659: nd_bpe = cbpe;
1.73 noro 2660: nd_setup_parameters(nd_nvar,0);
1.121 noro 2661: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2662: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2663: if ( nalg )
2664: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2665: MKLIST(*rp,cand);
1.1 noro 2666: }
2667:
1.61 noro 2668: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2669: {
1.61 noro 2670: UINT *d;
1.57 noro 2671: int i,j,l,s,ord_l;
1.43 noro 2672: struct order_pair *op;
1.1 noro 2673:
2674: d = dl->d;
1.41 noro 2675: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2676: if ( nd_blockmask ) {
2677: l = nd_blockmask->n;
2678: op = nd_blockmask->order_pair;
2679: for ( j = 0, s = 0; j < l; j++ ) {
2680: ord_l = op[j].length;
1.57 noro 2681: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2682: }
2683: TD(r) = ndl_weight(r);
1.61 noro 2684: ndl_weight_mask(r);
1.43 noro 2685: } else {
1.56 noro 2686: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2687: TD(r) = ndl_weight(r);
2688: }
1.1 noro 2689: }
2690:
1.61 noro 2691: DL ndltodl(int n,UINT *ndl)
1.1 noro 2692: {
2693: DL dl;
2694: int *d;
1.57 noro 2695: int i,j,l,s,ord_l;
1.43 noro 2696: struct order_pair *op;
1.1 noro 2697:
2698: NEWDL(dl,n);
1.34 noro 2699: dl->td = TD(ndl);
1.1 noro 2700: d = dl->d;
1.43 noro 2701: if ( nd_blockmask ) {
2702: l = nd_blockmask->n;
2703: op = nd_blockmask->order_pair;
2704: for ( j = 0, s = 0; j < l; j++ ) {
2705: ord_l = op[j].length;
1.57 noro 2706: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2707: }
2708: } else {
1.56 noro 2709: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2710: }
1.1 noro 2711: return dl;
2712: }
2713:
1.61 noro 2714: void ndl_print(UINT *dl)
1.1 noro 2715: {
2716: int n;
1.57 noro 2717: int i,j,l,ord_l,s,s0;
1.43 noro 2718: struct order_pair *op;
1.1 noro 2719:
2720: n = nd_nvar;
2721: printf("<<");
1.43 noro 2722: if ( nd_blockmask ) {
2723: l = nd_blockmask->n;
2724: op = nd_blockmask->order_pair;
2725: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2726: ord_l = op[j].length;
1.57 noro 2727: for ( i = 0; i < ord_l; i++, s++ )
2728: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2729: }
2730: } else {
1.56 noro 2731: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2732: }
1.1 noro 2733: printf(">>");
2734: }
2735:
2736: void nd_print(ND p)
2737: {
2738: NM m;
2739:
2740: if ( !p )
2741: printf("0\n");
2742: else {
2743: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2744: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2745: else printf("+%d*",CM(m));
1.14 noro 2746: ndl_print(DL(m));
1.1 noro 2747: }
2748: printf("\n");
2749: }
2750: }
2751:
1.113 noro 2752: void nd_print_q(ND p)
1.16 noro 2753: {
2754: NM m;
2755:
2756: if ( !p )
2757: printf("0\n");
2758: else {
2759: for ( m = BDY(p); m; m = NEXT(m) ) {
2760: printf("+");
1.113 noro 2761: printexpr(CO,(Obj)CQ(m));
1.16 noro 2762: printf("*");
2763: ndl_print(DL(m));
2764: }
2765: printf("\n");
2766: }
2767: }
2768:
1.1 noro 2769: void ndp_print(ND_pairs d)
2770: {
2771: ND_pairs t;
2772:
1.34 noro 2773: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2774: printf("\n");
2775: }
2776:
1.20 noro 2777: void nd_removecont(int mod,ND p)
1.16 noro 2778: {
2779: int i,n;
1.113 noro 2780: Q *w;
2781: Q dvr,t;
1.16 noro 2782: NM m;
1.21 noro 2783: struct oVECT v;
1.113 noro 2784: N q,r;
1.16 noro 2785:
1.71 noro 2786: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2787: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2788: else {
2789: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2790: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2791: v.len = n;
2792: v.body = (pointer *)w;
1.113 noro 2793: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2794: removecont_array(w,n);
1.113 noro 2795: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2796: }
2797: }
2798:
1.21 noro 2799: void nd_removecont2(ND p1,ND p2)
2800: {
2801: int i,n1,n2,n;
1.113 noro 2802: Q *w;
2803: Q dvr,t;
1.21 noro 2804: NM m;
2805: struct oVECT v;
1.113 noro 2806: N q,r;
1.21 noro 2807:
2808: if ( !p1 ) {
2809: nd_removecont(0,p2); return;
2810: } else if ( !p2 ) {
2811: nd_removecont(0,p1); return;
2812: }
2813: n1 = nd_length(p1);
2814: n2 = nd_length(p2);
2815: n = n1+n2;
1.113 noro 2816: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2817: v.len = n;
2818: v.body = (pointer *)w;
1.113 noro 2819: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2820: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2821: removecont_array(w,n);
1.113 noro 2822: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2823: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2824: }
2825:
1.20 noro 2826: void ndv_removecont(int mod,NDV p)
1.16 noro 2827: {
2828: int i,len;
1.113 noro 2829: Q *w;
2830: Q dvr,t;
1.16 noro 2831: NMV m;
2832:
1.71 noro 2833: if ( mod == -1 )
2834: ndv_mul_c(mod,p,_invsf(HCM(p)));
2835: else if ( mod )
1.20 noro 2836: ndv_mul_c(mod,p,invm(HCM(p),mod));
2837: else {
2838: len = p->len;
1.113 noro 2839: w = (Q *)ALLOCA(len*sizeof(Q));
2840: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
2841: sortbynm(w,len);
2842: qltozl(w,len,&dvr);
2843: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2844: divq(CQ(m),dvr,&t); CQ(m) = t;
2845: }
1.16 noro 2846: }
1.21 noro 2847: }
2848:
1.61 noro 2849: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2850: {
2851: int len,i,max;
2852: NMV m,mr0,mr,t;
2853:
2854: len = p->len;
2855: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2856: max = MAX(max,TD(DL(m)));
2857: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2858: m = (NMV)((char *)mr0+(len-1)*oadv);
2859: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2860: t = (NMV)ALLOCA(nmv_adv);
2861: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2862: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2863: CQ(mr) = CQ(m);
1.61 noro 2864: ndl_copy(DL(t),DL(mr));
2865: }
2866: NV(p)++;
2867: BDY(p) = mr0;
2868: }
2869:
1.45 noro 2870: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2871: {
1.45 noro 2872: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2873: Q *w;
2874: Q dvr,t;
1.23 noro 2875: NMV m,r;
2876:
2877: len = p->len;
2878: newnvar = nd_nvar-1;
1.48 noro 2879: newexporigin = nd_get_exporigin(ord);
1.45 noro 2880: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2881: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2882: ndl_dehomogenize(DL(m));
1.23 noro 2883: if ( newwpd != nd_wpd ) {
1.90 noro 2884: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2885: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2886: CQ(r) = CQ(m);
1.45 noro 2887: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2888: adj = nd_exporigin-newexporigin;
2889: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2890: }
2891: }
2892: NV(p)--;
2893: }
2894:
1.113 noro 2895: void removecont_array(Q *c,int n)
1.21 noro 2896: {
2897: struct oVECT v;
1.113 noro 2898: Q d0,d1,a,u,u1,gcd;
1.110 noro 2899: int i,j;
1.113 noro 2900: N qn,rn,gn;
2901: Q *q,*r;
1.21 noro 2902:
1.113 noro 2903: q = (Q *)ALLOCA(n*sizeof(Q));
2904: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2905: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2906: igcdv_estimate(&v,&d0);
2907: for ( i = 0; i < n; i++ ) {
2908: divn(NM(c[i]),NM(d0),&qn,&rn);
2909: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2910: NTOQ(rn,SGN(c[i]),r[i]);
2911: }
1.34 noro 2912: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2913: if ( i < n ) {
2914: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2915: igcdv(&v,&d1);
2916: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2917: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2918: for ( i = 0; i < n; i++ ) {
1.113 noro 2919: mulq(a,q[i],&u);
2920: if ( r[i] ) {
2921: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2922: addq(u,u1,&q[i]);
2923: } else
2924: q[i] = u;
1.21 noro 2925: }
2926: }
1.34 noro 2927: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2928: }
2929:
1.19 noro 2930: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2931: {
2932: NM m;
2933: int c,c1;
2934:
1.34 noro 2935: if ( !p ) return;
1.115 noro 2936: if ( mul == 1 ) return;
1.71 noro 2937: if ( mod == -1 )
2938: for ( m = BDY(p); m; m = NEXT(m) )
2939: CM(m) = _mulsf(CM(m),mul);
2940: else
2941: for ( m = BDY(p); m; m = NEXT(m) ) {
2942: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2943: }
1.1 noro 2944: }
2945:
1.113 noro 2946: void nd_mul_c_q(ND p,Q mul)
1.16 noro 2947: {
2948: NM m;
1.113 noro 2949: Q c;
1.16 noro 2950:
1.113 noro 2951: if ( !p ) return;
1.115 noro 2952: if ( UNIQ(mul) ) return;
1.16 noro 2953: for ( m = BDY(p); m; m = NEXT(m) ) {
1.113 noro 2954: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 2955: }
2956: }
2957:
1.61 noro 2958: void nd_mul_c_p(VL vl,ND p,P mul)
2959: {
2960: NM m;
2961: P c;
2962:
2963: if ( !p ) return;
2964: for ( m = BDY(p); m; m = NEXT(m) ) {
2965: mulp(vl,CP(m),mul,&c); CP(m) = c;
2966: }
2967: }
2968:
1.1 noro 2969: void nd_free(ND p)
2970: {
2971: NM t,s;
2972:
1.34 noro 2973: if ( !p ) return;
1.1 noro 2974: t = BDY(p);
2975: while ( t ) {
2976: s = NEXT(t);
2977: FREENM(t);
2978: t = s;
2979: }
2980: FREEND(p);
2981: }
2982:
1.23 noro 2983: void ndv_free(NDV p)
2984: {
2985: GC_free(BDY(p));
2986: }
2987:
1.61 noro 2988: void nd_append_red(UINT *d,int i)
1.1 noro 2989: {
1.13 noro 2990: RHist m,m0;
1.1 noro 2991: int h;
2992:
1.13 noro 2993: NEWRHist(m);
1.34 noro 2994: h = ndl_hash_value(d);
1.13 noro 2995: m->index = i;
1.14 noro 2996: ndl_copy(d,DL(m));
1.1 noro 2997: NEXT(m) = nd_red[h];
2998: nd_red[h] = m;
2999: }
3000:
1.61 noro 3001: UINT *ndv_compute_bound(NDV p)
1.1 noro 3002: {
1.61 noro 3003: UINT *d1,*d2,*t;
3004: UINT u;
1.57 noro 3005: int i,j,k,l,len,ind;
1.45 noro 3006: NMV m;
1.1 noro 3007:
3008: if ( !p )
3009: return 0;
1.61 noro 3010: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3011: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 3012: len = LEN(p);
3013: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3014: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3015: ndl_lcm(DL(m),d1,d2);
1.1 noro 3016: t = d1; d1 = d2; d2 = t;
3017: }
1.12 noro 3018: l = nd_nvar+31;
1.61 noro 3019: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3020: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3021: u = d1[i];
3022: k = (nd_epw-1)*nd_bpe;
3023: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3024: t[ind] = (u>>k)&nd_mask0;
3025: }
3026: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3027: return t;
3028: }
3029:
1.99 noro 3030: UINT *nd_compute_bound(ND p)
3031: {
3032: UINT *d1,*d2,*t;
3033: UINT u;
3034: int i,j,k,l,len,ind;
3035: NM m;
3036:
3037: if ( !p )
3038: return 0;
3039: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3040: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3041: len = LEN(p);
3042: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3043: for ( m = NEXT(m); m; m = NEXT(m) ) {
3044: ndl_lcm(DL(m),d1,d2);
3045: t = d1; d1 = d2; d2 = t;
3046: }
3047: l = nd_nvar+31;
3048: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3049: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3050: u = d1[i];
3051: k = (nd_epw-1)*nd_bpe;
3052: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3053: t[ind] = (u>>k)&nd_mask0;
3054: }
3055: for ( ; ind < l; ind++ ) t[ind] = 0;
3056: return t;
3057: }
3058:
1.48 noro 3059: int nd_get_exporigin(struct order_spec *ord)
3060: {
1.51 noro 3061: switch ( ord->id ) {
1.96 noro 3062: case 0: case 2:
1.48 noro 3063: return 1;
1.41 noro 3064: case 1:
3065: /* block order */
1.43 noro 3066: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3067: return ord->ord.block.length+1;
1.96 noro 3068: case 3:
3069: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3070: }
1.48 noro 3071: }
3072:
1.61 noro 3073: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3074: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3075: struct order_pair *op;
1.48 noro 3076:
1.73 noro 3077: nd_nvar = nvar;
1.79 noro 3078: if ( max ) {
1.83 noro 3079: /* XXX */
3080: if ( do_weyl ) nd_bpe = 32;
3081: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3082: else if ( max < 4 ) nd_bpe = 2;
3083: else if ( max < 8 ) nd_bpe = 3;
3084: else if ( max < 16 ) nd_bpe = 4;
3085: else if ( max < 32 ) nd_bpe = 5;
3086: else if ( max < 64 ) nd_bpe = 6;
3087: else if ( max < 256 ) nd_bpe = 8;
3088: else if ( max < 1024 ) nd_bpe = 10;
3089: else if ( max < 65536 ) nd_bpe = 16;
3090: else nd_bpe = 32;
3091: }
1.61 noro 3092: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3093: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3094: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3095: wpd = nd_exporigin+elen;
3096: if ( wpd != nd_wpd ) {
3097: nd_free_private_storage();
3098: nd_wpd = wpd;
3099: }
1.1 noro 3100: if ( nd_bpe < 32 ) {
3101: nd_mask0 = (1<<nd_bpe)-1;
3102: } else {
3103: nd_mask0 = 0xffffffff;
3104: }
3105: bzero(nd_mask,sizeof(nd_mask));
3106: nd_mask1 = 0;
3107: for ( i = 0; i < nd_epw; i++ ) {
3108: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3109: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3110: }
1.90 noro 3111: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3112: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3113: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3114: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3115: }
3116:
1.103 noro 3117: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3118: {
1.37 noro 3119: int i,obpe,oadv,h;
1.71 noro 3120: static NM prev_nm_free_list;
3121: static ND_pairs prev_ndp_free_list;
1.13 noro 3122: RHist mr0,mr;
3123: RHist r;
1.37 noro 3124: RHist *old_red;
1.71 noro 3125: ND_pairs s0,s,t;
1.43 noro 3126: EPOS oepos;
1.15 noro 3127:
1.1 noro 3128: obpe = nd_bpe;
1.11 noro 3129: oadv = nmv_adv;
1.43 noro 3130: oepos = nd_epos;
1.63 noro 3131: if ( obpe < 2 ) nd_bpe = 2;
3132: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3133: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3134: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3135: else if ( obpe < 6 ) nd_bpe = 6;
3136: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3137: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3138: else if ( obpe < 16 ) nd_bpe = 16;
3139: else if ( obpe < 32 ) nd_bpe = 32;
3140: else error("nd_reconstruct : exponent too large");
1.5 noro 3141:
1.73 noro 3142: nd_setup_parameters(nd_nvar,0);
1.1 noro 3143: prev_nm_free_list = _nm_free_list;
3144: prev_ndp_free_list = _ndp_free_list;
3145: _nm_free_list = 0;
3146: _ndp_free_list = 0;
1.53 noro 3147: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3148: if ( trace )
3149: for ( i = nd_psn-1; i >= 0; i-- )
3150: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3151: s0 = 0;
3152: for ( t = d; t; t = NEXT(t) ) {
3153: NEXTND_pairs(s0,s);
3154: s->i1 = t->i1;
3155: s->i2 = t->i2;
1.14 noro 3156: SG(s) = SG(t);
1.61 noro 3157: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3158: }
1.37 noro 3159:
3160: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3161: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3162: old_red[i] = nd_red[i];
3163: nd_red[i] = 0;
3164: }
3165: for ( i = 0; i < REDTAB_LEN; i++ )
3166: for ( r = old_red[i]; r; r = NEXT(r) ) {
3167: NEWRHist(mr);
1.13 noro 3168: mr->index = r->index;
1.20 noro 3169: SG(mr) = SG(r);
1.61 noro 3170: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3171: h = ndl_hash_value(DL(mr));
3172: NEXT(mr) = nd_red[h];
3173: nd_red[h] = mr;
1.6 noro 3174: }
1.37 noro 3175: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3176: old_red = 0;
1.11 noro 3177: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3178: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3179: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3180: nd_psh[i] = r;
1.11 noro 3181: }
1.1 noro 3182: if ( s0 ) NEXT(s) = 0;
3183: prev_nm_free_list = 0;
3184: prev_ndp_free_list = 0;
1.71 noro 3185: #if 0
1.1 noro 3186: GC_gcollect();
1.71 noro 3187: #endif
1.1 noro 3188: return s0;
3189: }
3190:
1.61 noro 3191: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3192: {
1.57 noro 3193: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3194: struct order_pair *op;
1.1 noro 3195:
3196: n = nd_nvar;
1.61 noro 3197: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3198: omask0 = (1<<obpe)-1;
1.34 noro 3199: TD(r) = TD(d);
1.41 noro 3200: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3201: if ( nd_blockmask ) {
3202: l = nd_blockmask->n;
3203: op = nd_blockmask->order_pair;
3204: for ( i = 1; i < nd_exporigin; i++ )
3205: r[i] = d[i];
3206: for ( j = 0, s = 0; j < l; j++ ) {
3207: ord_l = op[j].length;
1.57 noro 3208: for ( i = 0; i < ord_l; i++, s++ ) {
3209: ei = GET_EXP_OLD(d,s);
3210: PUT_EXP(r,s,ei);
3211: }
1.1 noro 3212: }
1.43 noro 3213: } else {
1.56 noro 3214: for ( i = 0; i < n; i++ ) {
3215: ei = GET_EXP_OLD(d,i);
3216: PUT_EXP(r,i,ei);
3217: }
1.1 noro 3218: }
3219: }
1.3 noro 3220:
1.6 noro 3221: ND nd_copy(ND p)
3222: {
3223: NM m,mr,mr0;
1.41 noro 3224: int c,n;
1.6 noro 3225: ND r;
3226:
3227: if ( !p )
3228: return 0;
3229: else {
3230: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3231: NEXTNM(mr0,mr);
1.14 noro 3232: CM(mr) = CM(m);
3233: ndl_copy(DL(m),DL(mr));
1.6 noro 3234: }
3235: NEXT(mr) = 0;
1.31 noro 3236: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3237: SG(r) = SG(p);
1.6 noro 3238: return r;
3239: }
3240: }
3241:
1.53 noro 3242: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3243: {
3244: NM m;
3245: NDV p1,p2;
3246: ND t1,t2;
1.61 noro 3247: UINT *lcm;
1.31 noro 3248: int td;
1.11 noro 3249:
1.77 noro 3250: if ( !mod && nd_demand ) {
3251: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3252: } else {
1.77 noro 3253: if ( trace ) {
3254: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3255: } else {
3256: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3257: }
1.20 noro 3258: }
1.34 noro 3259: lcm = LCM(p);
1.11 noro 3260: NEWNM(m);
1.113 noro 3261: CQ(m) = HCQ(p2);
1.34 noro 3262: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3263: if ( ndl_check_bound2(p->i1,DL(m)) )
3264: return 0;
1.55 noro 3265: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3266: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3267: else if ( mod ) CM(m) = mod-HCM(p1);
1.113 noro 3268: else chsgnq(HCQ(p1),&CQ(m));
1.34 noro 3269: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3270: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3271: nd_free(t1);
3272: return 0;
3273: }
1.55 noro 3274: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3275: *rp = nd_add(mod,t1,t2);
1.11 noro 3276: FREENM(m);
3277: return 1;
3278: }
3279:
1.19 noro 3280: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3281: {
3282: NMV m;
3283: int c,c1,len,i;
3284:
1.34 noro 3285: if ( !p ) return;
1.14 noro 3286: len = LEN(p);
1.71 noro 3287: if ( mod == -1 )
3288: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3289: CM(m) = _mulsf(CM(m),mul);
3290: else
3291: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3292: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3293: }
1.11 noro 3294: }
3295:
1.113 noro 3296: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3297: {
3298: NMV m;
1.113 noro 3299: Q c;
1.16 noro 3300: int len,i;
3301:
1.34 noro 3302: if ( !p ) return;
1.16 noro 3303: len = LEN(p);
3304: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3305: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3306: }
3307: }
3308:
1.55 noro 3309: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3310: int n2,i,j,l,n,tlen;
1.61 noro 3311: UINT *d0;
1.55 noro 3312: NM *tab,*psum;
3313: ND s,r;
3314: NM t;
3315: NMV m1;
3316:
3317: if ( !p ) return 0;
3318: n = NV(p); n2 = n>>1;
3319: d0 = DL(m0);
3320: l = LEN(p);
3321: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3322: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3323: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3324: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3325: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3326: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3327: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3328: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3329: for ( j = 0; j < tlen; j++ ) {
3330: if ( tab[j] ) {
3331: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3332: }
3333: }
3334: }
3335: for ( i = tlen-1, r = 0; i >= 0; i-- )
3336: if ( psum[i] ) {
3337: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3338: MKND(n,psum[i],j,s);
3339: r = nd_add(mod,r,s);
3340: }
1.56 noro 3341: if ( r ) SG(r) = SG(p)+TD(d0);
3342: return r;
1.55 noro 3343: }
3344:
1.56 noro 3345: /* product of monomials */
3346: /* XXX block order is not handled correctly */
3347:
1.55 noro 3348: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3349: {
1.56 noro 3350: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3351: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3352: Q *ctab_q;
3353: Q q,q1;
1.61 noro 3354: UINT c0,c1,c;
1.55 noro 3355: NM *p;
3356: NM m,t;
3357:
3358: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3359: if ( !m0 || !m1 ) return;
3360: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3361: NEWNM(m); d = DL(m);
1.56 noro 3362: if ( mod ) {
3363: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3364: } else
1.113 noro 3365: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3366: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3367: homo = n&1 ? 1 : 0;
3368: if ( homo ) {
3369: /* offset of h-degree */
3370: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3371: PUT_EXP(DL(m),n-1,h);
3372: TD(DL(m)) = h;
1.61 noro 3373: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3374: }
3375: tab[0] = m;
3376: NEWNM(m); d = DL(m);
1.57 noro 3377: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3378: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3379: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3380: /* xi^a*(Di^k*xi^l)*Di^b */
3381: a += l; b += k;
1.56 noro 3382: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3383: if ( !k || !l ) {
3384: for ( j = 0; j < curlen; j++ )
1.56 noro 3385: if ( t = tab[j] ) {
3386: dt = DL(t);
3387: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3388: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3389: }
3390: curlen *= k+1;
3391: continue;
3392: }
3393: min = MIN(k,l);
1.56 noro 3394: if ( mod ) {
1.61 noro 3395: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3396: mkwcm(k,l,mod,ctab);
3397: } else {
1.113 noro 3398: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3399: mkwc(k,l,ctab_q);
1.56 noro 3400: }
1.57 noro 3401: for ( j = min; j >= 0; j-- ) {
1.56 noro 3402: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3403: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3404: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3405: if ( homo ) {
3406: TD(d) = s;
1.56 noro 3407: PUT_EXP(d,n-1,s-h);
1.55 noro 3408: } else TD(d) = h;
1.61 noro 3409: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3410: if ( mod ) c = ctab[j];
1.113 noro 3411: else q = ctab_q[j];
1.57 noro 3412: p = tab+curlen*j;
3413: if ( j == 0 ) {
3414: for ( u = 0; u < curlen; u++, p++ ) {
3415: if ( tab[u] ) {
3416: ndl_addto(DL(tab[u]),d);
3417: if ( mod ) {
3418: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3419: } else {
1.113 noro 3420: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3421: }
3422: }
1.56 noro 3423: }
1.57 noro 3424: } else {
3425: for ( u = 0; u < curlen; u++, p++ ) {
3426: if ( tab[u] ) {
3427: NEWNM(t);
3428: ndl_add(DL(tab[u]),d,DL(t));
3429: if ( mod ) {
3430: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3431: } else
1.113 noro 3432: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3433: *p = t;
3434: }
1.55 noro 3435: }
3436: }
3437: }
3438: curlen *= k+1;
3439: }
3440: FREENM(m);
3441: }
3442:
1.63 noro 3443: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3444: {
3445: NM mr,mr0;
3446: NMV m;
3447: UINT *d,*dt,*dm;
3448: int c,n,td,i,c1,c2,len;
1.113 noro 3449: Q q;
1.63 noro 3450: ND r;
3451:
3452: if ( !p ) return 0;
3453: else {
3454: n = NV(p); m = BDY(p);
3455: d = DL(m0);
3456: len = LEN(p);
3457: mr0 = 0;
3458: td = TD(d);
3459: c = CM(m0);
3460: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3461: NEXTNM(mr0,mr);
3462: CM(mr) = 1;
3463: ndl_add(DL(m),d,DL(mr));
3464: }
3465: NEXT(mr) = 0;
3466: MKND(NV(p),mr0,len,r);
3467: SG(r) = SG(p) + TD(d);
3468: return r;
3469: }
3470: }
3471:
1.55 noro 3472: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3473: {
3474: NM mr,mr0;
3475: NMV m;
1.61 noro 3476: UINT *d,*dt,*dm;
1.9 noro 3477: int c,n,td,i,c1,c2,len;
1.113 noro 3478: Q q;
1.9 noro 3479: ND r;
3480:
1.34 noro 3481: if ( !p ) return 0;
1.55 noro 3482: else if ( do_weyl )
1.71 noro 3483: if ( mod == -1 )
3484: error("ndv_mul_nm : not implemented (weyl)");
3485: else
3486: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3487: else {
3488: n = NV(p); m = BDY(p);
1.34 noro 3489: d = DL(m0);
1.14 noro 3490: len = LEN(p);
1.9 noro 3491: mr0 = 0;
1.34 noro 3492: td = TD(d);
1.71 noro 3493: if ( mod == -1 ) {
3494: c = CM(m0);
3495: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3496: NEXTNM(mr0,mr);
3497: CM(mr) = _mulsf(CM(m),c);
3498: ndl_add(DL(m),d,DL(mr));
3499: }
3500: } else if ( mod ) {
1.16 noro 3501: c = CM(m0);
3502: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3503: NEXTNM(mr0,mr);
3504: c1 = CM(m);
1.19 noro 3505: DMAR(c1,c,0,mod,c2);
1.16 noro 3506: CM(mr) = c2;
3507: ndl_add(DL(m),d,DL(mr));
3508: }
3509: } else {
1.113 noro 3510: q = CQ(m0);
1.16 noro 3511: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3512: NEXTNM(mr0,mr);
1.113 noro 3513: mulq(CQ(m),q,&CQ(mr));
1.16 noro 3514: ndl_add(DL(m),d,DL(mr));
3515: }
1.4 noro 3516: }
1.9 noro 3517: NEXT(mr) = 0;
1.31 noro 3518: MKND(NV(p),mr0,len,r);
1.34 noro 3519: SG(r) = SG(p) + TD(d);
1.9 noro 3520: return r;
1.4 noro 3521: }
3522: }
3523:
1.104 noro 3524: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3525: {
3526: NM mq0,mq;
1.102 noro 3527: NMV tm;
1.113 noro 3528: Q q;
1.104 noro 3529: int i,nv,sg,c,c1,c2,hindex;
3530: ND p,t,r;
3531: N tnm;
3532:
1.124 noro 3533: if ( bucket->m < 0 ) return 0;
1.99 noro 3534: else {
1.104 noro 3535: nv = NV(d);
1.99 noro 3536: mq0 = 0;
1.102 noro 3537: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3538: while ( 1 ) {
1.113 noro 3539: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3540: if ( hindex < 0 ) break;
3541: p = bucket->body[hindex];
1.99 noro 3542: NEXTNM(mq0,mq);
3543: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3544: ndl_copy(DL(mq),DL(tm));
3545: if ( mod ) {
3546: c1 = invm(HCM(d),mod); c2 = HCM(p);
3547: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3548: CM(tm) = mod-c;
3549: } else {
1.113 noro 3550: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3551: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3552: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3553: }
3554: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3555: bucket->body[hindex] = nd_remove_head(p);
3556: t = nd_remove_head(t);
3557: add_pbucket(mod,bucket,t);
3558: }
3559: if ( !mq0 )
3560: r = 0;
3561: else {
3562: NEXT(mq) = 0;
3563: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3564: MKND(nv,mq0,i,r);
3565: /* XXX */
3566: SG(r) = HTD(r);
1.99 noro 3567: }
3568: return r;
3569: }
3570: }
3571:
1.43 noro 3572: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3573: {
1.13 noro 3574: NMV m,mr,mr0,t;
3575: int len,i,k;
1.11 noro 3576:
1.61 noro 3577: if ( !p ) return;
3578: m = BDY(p); len = LEN(p);
3579: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3580: m = (NMV)((char *)mr0+(len-1)*oadv);
3581: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3582: t = (NMV)ALLOCA(nmv_adv);
3583: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3584: CQ(t) = CQ(m);
1.61 noro 3585: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3586: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3587: CQ(mr) = CQ(t);
1.61 noro 3588: ndl_copy(DL(t),DL(mr));
3589: }
3590: BDY(p) = mr0;
3591: }
3592:
3593: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3594: {
3595: NMV m,mr,mr0;
3596: int len,i;
3597: NDV r;
1.11 noro 3598:
1.61 noro 3599: if ( !p ) return 0;
3600: m = BDY(p); len = LEN(p);
3601: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3602: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3603: ndl_zero(DL(mr));
3604: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3605: CQ(mr) = CQ(m);
1.11 noro 3606: }
1.61 noro 3607: MKNDV(NV(p),mr0,len,r);
3608: SG(r) = SG(p);
3609: return r;
1.11 noro 3610: }
3611:
1.61 noro 3612: /* duplicate p */
3613:
3614: NDV ndv_dup(int mod,NDV p)
1.3 noro 3615: {
3616: NDV d;
1.61 noro 3617: NMV t,m,m0;
1.3 noro 3618: int i,len;
3619:
1.34 noro 3620: if ( !p ) return 0;
1.31 noro 3621: len = LEN(p);
1.34 noro 3622: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3623: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3624: ndl_copy(DL(t),DL(m));
1.113 noro 3625: CQ(m) = CQ(t);
1.3 noro 3626: }
3627: MKNDV(NV(p),m0,len,d);
1.23 noro 3628: SG(d) = SG(p);
3629: return d;
3630: }
3631:
1.63 noro 3632: ND nd_dup(ND p)
3633: {
3634: ND d;
3635: NM t,m,m0;
3636:
3637: if ( !p ) return 0;
3638: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3639: NEXTNM(m0,m);
3640: ndl_copy(DL(t),DL(m));
1.113 noro 3641: CQ(m) = CQ(t);
1.63 noro 3642: }
3643: if ( m0 ) NEXT(m) = 0;
3644: MKND(NV(p),m0,LEN(p),d);
3645: SG(d) = SG(p);
3646: return d;
3647: }
3648:
1.61 noro 3649: /* XXX if p->len == 0 then it represents 0 */
3650:
3651: void ndv_mod(int mod,NDV p)
3652: {
3653: NMV t,d;
1.125 noro 3654: int r,s,u;
1.61 noro 3655: int i,len,dlen;
1.71 noro 3656: Obj gfs;
1.61 noro 3657:
3658: if ( !p ) return;
3659: len = LEN(p);
3660: dlen = 0;
1.71 noro 3661: if ( mod == -1 )
3662: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3663: simp_ff((Obj)CP(t),&gfs);
3664: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3665: CM(d) = r;
3666: ndl_copy(DL(t),DL(d));
3667: NMV_ADV(d);
3668: dlen++;
3669: }
1.71 noro 3670: else
3671: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.113 noro 3672: r = rem(NM(CQ(t)),mod);
1.71 noro 3673: if ( r ) {
1.113 noro 3674: if ( SGN(CQ(t)) < 0 )
3675: r = mod-r;
1.125 noro 3676: if ( DN(CQ(t)) ) {
3677: s = rem(DN(CQ(t)),mod);
3678: if ( !s )
3679: error("ndv_mod : division by 0");
3680: s = invm(s,mod);
3681: DMAR(r,s,0,mod,u); r = u;
3682: }
1.71 noro 3683: CM(d) = r;
3684: ndl_copy(DL(t),DL(d));
3685: NMV_ADV(d);
3686: dlen++;
3687: }
3688: }
1.61 noro 3689: LEN(p) = dlen;
3690: }
3691:
3692: NDV ptondv(VL vl,VL dvl,P p)
3693: {
3694: ND nd;
3695:
3696: nd = ptond(vl,dvl,p);
3697: return ndtondv(0,nd);
3698: }
3699:
3700: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3701: {
1.61 noro 3702: int n,i,j,k,e;
3703: VL tvl;
3704: V v;
3705: DCP dc;
3706: DCP *w;
3707: ND r,s,t,u;
3708: P x;
3709: int c;
3710: UINT *d;
1.23 noro 3711: NM m,m0;
1.61 noro 3712:
3713: if ( !p )
3714: return 0;
3715: else if ( NUM(p) ) {
3716: NEWNM(m);
3717: ndl_zero(DL(m));
1.113 noro 3718: CQ(m) = (Q)p;
1.61 noro 3719: NEXT(m) = 0;
3720: MKND(nd_nvar,m,1,r);
3721: SG(r) = 0;
3722: return r;
3723: } else {
3724: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3725: w = (DCP *)ALLOCA(k*sizeof(DCP));
3726: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3727: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3728: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3729: if ( !tvl ) {
3730: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3731: t = ptond(vl,dvl,COEF(w[j]));
3732: pwrp(vl,x,DEG(w[j]),&p);
3733: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3734: }
3735: return s;
3736: } else {
3737: NEWNM(m0); d = DL(m0);
3738: for ( j = k-1, s = 0; j >= 0; j-- ) {
3739: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3740: TD(d) = MUL_WEIGHT(e,i);
3741: if ( nd_blockmask) ndl_weight_mask(d);
3742: t = ptond(vl,dvl,COEF(w[j]));
3743: for ( m = BDY(t); m; m = NEXT(m) )
3744: ndl_addto(DL(m),d);
3745: SG(t) += TD(d);
3746: s = nd_add(0,s,t);
3747: }
3748: FREENM(m0);
3749: return s;
3750: }
3751: }
3752: }
3753:
3754: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3755: {
3756: VL tvl;
3757: int len,n,j,i,e;
3758: NMV m;
3759: Q q;
3760: P c;
3761: UINT *d;
3762: P s,r,u,t,w;
1.71 noro 3763: GFS gfs;
1.23 noro 3764:
1.34 noro 3765: if ( !p ) return 0;
1.61 noro 3766: else {
3767: len = LEN(p);
3768: n = NV(p);
3769: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3770: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3771: if ( mod == -1 ) {
3772: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3773: } else if ( mod ) {
1.61 noro 3774: STOQ(CM(m),q); c = (P)q;
3775: } else
1.113 noro 3776: c = CP(m);
1.61 noro 3777: d = DL(m);
3778: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3779: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3780: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3781: }
3782: addp(vl,s,t,&u); s = u;
3783: }
3784: return s;
1.23 noro 3785: }
1.3 noro 3786: }
3787:
1.61 noro 3788: NDV ndtondv(int mod,ND p)
1.11 noro 3789: {
3790: NDV d;
1.61 noro 3791: NMV m,m0;
3792: NM t;
3793: int i,len;
1.11 noro 3794:
1.34 noro 3795: if ( !p ) return 0;
1.61 noro 3796: len = LEN(p);
1.74 noro 3797: if ( mod )
3798: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3799: else
3800: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3801: #if 0
1.74 noro 3802: ndv_alloc += nmv_adv*len;
1.103 noro 3803: #endif
1.61 noro 3804: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3805: ndl_copy(DL(t),DL(m));
1.113 noro 3806: CQ(m) = CQ(t);
1.11 noro 3807: }
1.61 noro 3808: MKNDV(NV(p),m0,len,d);
1.14 noro 3809: SG(d) = SG(p);
1.11 noro 3810: return d;
3811: }
3812:
1.61 noro 3813: ND ndvtond(int mod,NDV p)
1.11 noro 3814: {
1.61 noro 3815: ND d;
3816: NM m,m0;
1.11 noro 3817: NMV t;
1.61 noro 3818: int i,len;
1.11 noro 3819:
1.34 noro 3820: if ( !p ) return 0;
1.11 noro 3821: m0 = 0;
1.61 noro 3822: len = p->len;
3823: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3824: NEXTNM(m0,m);
3825: ndl_copy(DL(t),DL(m));
1.113 noro 3826: CQ(m) = CQ(t);
1.11 noro 3827: }
3828: NEXT(m) = 0;
1.61 noro 3829: MKND(NV(p),m0,len,d);
1.14 noro 3830: SG(d) = SG(p);
1.11 noro 3831: return d;
3832: }
3833:
1.3 noro 3834: void ndv_print(NDV p)
3835: {
3836: NMV m;
3837: int i,len;
3838:
1.34 noro 3839: if ( !p ) printf("0\n");
1.3 noro 3840: else {
1.14 noro 3841: len = LEN(p);
1.3 noro 3842: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3843: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3844: else printf("+%d*",CM(m));
1.16 noro 3845: ndl_print(DL(m));
3846: }
3847: printf("\n");
3848: }
3849: }
3850:
1.113 noro 3851: void ndv_print_q(NDV p)
1.16 noro 3852: {
3853: NMV m;
3854: int i,len;
3855:
1.34 noro 3856: if ( !p ) printf("0\n");
1.16 noro 3857: else {
3858: len = LEN(p);
3859: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3860: printf("+");
1.113 noro 3861: printexpr(CO,(Obj)CQ(m));
1.16 noro 3862: printf("*");
1.14 noro 3863: ndl_print(DL(m));
1.3 noro 3864: }
3865: printf("\n");
3866: }
1.25 noro 3867: }
3868:
1.61 noro 3869: NODE ndv_reducebase(NODE x)
1.27 noro 3870: {
3871: int len,i,j;
3872: NDV *w;
3873: NODE t,t0;
3874:
3875: len = length(x);
3876: w = (NDV *)ALLOCA(len*sizeof(NDV));
3877: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3878: for ( i = 0; i < len; i++ ) {
3879: for ( j = 0; j < i; j++ ) {
3880: if ( w[i] && w[j] )
3881: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3882: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3883: }
3884: }
3885: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3886: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3887: }
3888: NEXT(t) = 0; x = t0;
3889: return x;
1.11 noro 3890: }
1.32 noro 3891:
1.43 noro 3892: /* XXX incomplete */
3893:
1.32 noro 3894: void nd_init_ord(struct order_spec *ord)
3895: {
1.43 noro 3896: switch ( ord->id ) {
1.32 noro 3897: case 0:
1.43 noro 3898: switch ( ord->ord.simple ) {
3899: case 0:
3900: nd_dcomp = 1;
3901: nd_isrlex = 1;
3902: break;
3903: case 1:
3904: nd_dcomp = 1;
3905: nd_isrlex = 0;
3906: break;
3907: case 2:
3908: nd_dcomp = 0;
3909: nd_isrlex = 0;
1.45 noro 3910: ndl_compare_function = ndl_lex_compare;
1.58 noro 3911: break;
3912: case 11:
3913: /* XXX */
3914: nd_dcomp = 0;
3915: nd_isrlex = 1;
3916: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3917: break;
3918: default:
3919: error("nd_gr : unsupported order");
3920: }
1.32 noro 3921: break;
3922: case 1:
1.96 noro 3923: /* block order */
1.43 noro 3924: /* XXX */
3925: nd_dcomp = -1;
1.32 noro 3926: nd_isrlex = 0;
1.45 noro 3927: ndl_compare_function = ndl_block_compare;
1.34 noro 3928: break;
1.43 noro 3929: case 2:
1.96 noro 3930: /* matrix order */
3931: /* XXX */
3932: nd_dcomp = -1;
3933: nd_isrlex = 0;
3934: nd_matrix_len = ord->ord.matrix.row;
3935: nd_matrix = ord->ord.matrix.matrix;
3936: ndl_compare_function = ndl_matrix_compare;
3937: break;
3938: case 3:
1.97 noro 3939: /* composite order */
3940: nd_dcomp = -1;
3941: nd_isrlex = 0;
3942: nd_worb_len = ord->ord.composite.length;
3943: nd_worb = ord->ord.composite.w_or_b;
3944: ndl_compare_function = ndl_composite_compare;
1.32 noro 3945: break;
3946: }
1.41 noro 3947: nd_ord = ord;
1.32 noro 3948: }
3949:
1.43 noro 3950: BlockMask nd_create_blockmask(struct order_spec *ord)
3951: {
3952: int n,i,j,s,l;
1.61 noro 3953: UINT *t;
1.43 noro 3954: BlockMask bm;
3955:
1.96 noro 3956: /* we only create mask table for block order */
3957: if ( ord->id != 1 )
1.43 noro 3958: return 0;
3959: n = ord->ord.block.length;
3960: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3961: bm->n = n;
3962: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3963: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3964: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3965: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3966: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3967: l = bm->order_pair[i].length;
3968: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3969: }
3970: return bm;
1.57 noro 3971: }
3972:
3973: EPOS nd_create_epos(struct order_spec *ord)
3974: {
3975: int i,j,l,s,ord_l,ord_o;
3976: EPOS epos;
3977: struct order_pair *op;
3978:
3979: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3980: switch ( ord->id ) {
3981: case 0:
3982: if ( nd_isrlex ) {
3983: for ( i = 0; i < nd_nvar; i++ ) {
3984: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3985: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3986: }
3987: } else {
3988: for ( i = 0; i < nd_nvar; i++ ) {
3989: epos[i].i = nd_exporigin + i/nd_epw;
3990: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3991: }
3992: }
3993: break;
3994: case 1:
3995: /* block order */
3996: l = ord->ord.block.length;
3997: op = ord->ord.block.order_pair;
3998: for ( j = 0, s = 0; j < l; j++ ) {
3999: ord_o = op[j].order;
4000: ord_l = op[j].length;
4001: if ( !ord_o )
4002: for ( i = 0; i < ord_l; i++ ) {
4003: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
4004: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
4005: }
4006: else
4007: for ( i = 0; i < ord_l; i++ ) {
4008: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4009: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4010: }
4011: s += ord_l;
4012: }
4013: break;
4014: case 2:
1.96 noro 4015: /* matrix order */
4016: case 3:
4017: /* composite order */
4018: for ( i = 0; i < nd_nvar; i++ ) {
4019: epos[i].i = nd_exporigin + i/nd_epw;
4020: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4021: }
4022: break;
1.57 noro 4023: }
4024: return epos;
1.43 noro 4025: }
1.59 noro 4026:
4027: /* external interface */
4028:
4029: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4030: {
1.61 noro 4031: NODE t,in0,in;
1.59 noro 4032: ND nd,nf;
1.61 noro 4033: NDV ndv;
4034: VL vv,tv;
4035: int stat,nvar,max,e;
1.113 noro 4036: union oNDC dn;
1.59 noro 4037:
1.116 noro 4038: if ( !f ) {
4039: *rp = 0;
4040: return;
4041: }
1.59 noro 4042: pltovl(v,&vv);
1.74 noro 4043: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4044:
4045: /* get the degree bound */
4046: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4047: for ( tv = vv; tv; tv = NEXT(tv) ) {
4048: e = getdeg(tv->v,(P)BDY(t));
4049: max = MAX(e,max);
4050: }
4051: for ( tv = vv; tv; tv = NEXT(tv) ) {
4052: e = getdeg(tv->v,f);
4053: max = MAX(e,max);
4054: }
4055:
1.59 noro 4056: nd_init_ord(ord);
1.61 noro 4057: nd_setup_parameters(nvar,max);
4058:
4059: /* conversion to ndv */
4060: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4061: NEXTNODE(in0,in);
4062: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 4063: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4064: }
4065: NEXTNODE(in0,in);
4066: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4067: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4068: NEXT(in) = 0;
4069:
1.128 noro 4070: /* dont sort */
4071: ndv_setup(m,0,in0,1);
1.59 noro 4072: nd_psn--;
4073: nd_scale=2;
4074: while ( 1 ) {
4075: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 4076: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 4077: if ( !stat ) {
4078: nd_psn++;
1.103 noro 4079: nd_reconstruct(0,0);
1.59 noro 4080: nd_psn--;
4081: } else
4082: break;
4083: }
1.61 noro 4084: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4085: }
4086:
4087: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4088: {
4089: NM m;
4090: UINT *t,*s;
4091: int i;
4092:
4093: for ( i = 0; i < n; i++ ) r[i] = 0;
4094: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4095: t = DL(m);
4096: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4097: r[i] = CM(m);
4098: }
4099: for ( i = 0; !r[i]; i++ );
4100: return i;
4101: }
4102:
1.113 noro 4103: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4104: {
1.107 noro 4105: NM m;
1.74 noro 4106: UINT *t,*s;
1.107 noro 4107: int i;
1.74 noro 4108:
4109: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4110: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4111: t = DL(m);
4112: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4113: r[i] = CQ(m);
1.74 noro 4114: }
4115: for ( i = 0; !r[i]; i++ );
4116: return i;
4117: }
4118:
1.129 noro 4119: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4120: {
4121: NM m;
4122: NMV mr;
4123: UINT *d,*t,*s;
4124: NDV p;
4125: int i,j,len;
4126: Q *r;
4127:
4128: m = pair->mul;
4129: d = DL(m);
4130: p = nd_ps[pair->index];
4131: len = LEN(p);
4132: r = (Q *)CALLOC(n,sizeof(Q));
4133: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4134: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4135: ndl_add(d,DL(mr),t);
4136: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4137: r[i] = CQ(mr);
4138: }
4139: return r;
4140: }
4141:
1.67 noro 4142: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4143: {
4144: NM m;
4145: NMV mr;
4146: UINT *d,*t,*s;
4147: NDV p;
1.67 noro 4148: unsigned char *ivc;
4149: unsigned short *ivs;
1.81 noro 4150: UINT *v,*ivi,*s0v;
1.67 noro 4151: int i,j,len,prev,diff,cdiff;
4152: IndArray r;
1.64 noro 4153:
4154: m = pair->mul;
4155: d = DL(m);
4156: p = nd_ps[pair->index];
4157: len = LEN(p);
4158: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4159: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4160: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4161: ndl_add(d,DL(mr),t);
4162: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4163: v[j] = i;
4164: }
1.81 noro 4165: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4166: r->head = v[0];
4167: diff = 0;
4168: for ( i = 1; i < len; i++ ) {
4169: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4170: }
4171: if ( diff < 256 ) {
4172: r->width = 1;
4173: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4174: r->index.c = ivc;
4175: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4176: } else if ( diff < 65536 ) {
4177: r->width = 2;
4178: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4179: r->index.s = ivs;
4180: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4181: } else {
4182: r->width = 4;
4183: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4184: r->index.i = ivi;
4185: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4186: }
1.67 noro 4187: return r;
1.64 noro 4188: }
4189:
1.135 noro 4190: int compress_array(Q *svect,Q *cvect,int n)
4191: {
4192: int i,j;
4193:
4194: for ( i = j = 0; i < n; i++ )
4195: if ( svect[i] ) cvect[j++] = svect[i];
4196: return j;
4197: }
4198:
4199: void expand_array(Q *svect,Q *cvect,int n)
4200: {
4201: int i,j;
4202:
4203: for ( i = j = 0; j < n; i++ )
4204: if ( svect[i] ) svect[i] = cvect[j++];
4205: }
4206:
1.133 noro 4207: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4208: {
1.135 noro 4209: int i,j,k,len,pos,prev,nz;
1.113 noro 4210: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4211: IndArray ivect;
4212: unsigned char *ivc;
4213: unsigned short *ivs;
4214: unsigned int *ivi;
4215: NDV redv;
4216: NMV mr;
4217: NODE rp;
4218: int maxrs;
1.133 noro 4219: double hmag;
1.135 noro 4220: Q *cvect;
1.107 noro 4221:
4222: maxrs = 0;
1.134 noro 4223: for ( i = 0; i < col && !svect[i]; i++ );
4224: if ( i == col ) return maxrs;
4225: hmag = p_mag((P)svect[i])*nd_scale;
1.135 noro 4226: cvect = (Q *)ALLOCA(col*sizeof(Q));
1.107 noro 4227: for ( i = 0; i < nred; i++ ) {
4228: ivect = imat[i];
4229: k = ivect->head;
4230: if ( svect[k] ) {
4231: maxrs = MAX(maxrs,rp0[i]->sugar);
1.133 noro 4232: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
1.107 noro 4233: len = LEN(redv); mr = BDY(redv);
1.113 noro 4234: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4235: chsgnq(cs,&mcs);
4236: if ( !UNIQ(cr) ) {
4237: for ( j = 0; j < col; j++ ) {
4238: mulq(svect[j],cr,&c1); svect[j] = c1;
4239: }
4240: }
1.107 noro 4241: svect[k] = 0; prev = k;
4242: switch ( ivect->width ) {
4243: case 1:
4244: ivc = ivect->index.c;
4245: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4246: pos = prev+ivc[j]; prev = pos;
1.113 noro 4247: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4248: }
4249: break;
4250: case 2:
4251: ivs = ivect->index.s;
4252: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4253: pos = prev+ivs[j]; prev = pos;
1.113 noro 4254: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4255: }
4256: break;
4257: case 4:
4258: ivi = ivect->index.i;
4259: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4260: pos = prev+ivi[j]; prev = pos;
1.113 noro 4261: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4262: }
4263: break;
4264: }
1.134 noro 4265: for ( j = k+1; j < col && !svect[j]; j++ );
4266: if ( j == col ) break;
4267: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
1.135 noro 4268: nz = compress_array(svect,cvect,col);
4269: removecont_array(cvect,nz);
4270: expand_array(svect,cvect,nz);
1.134 noro 4271: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
4272: }
1.107 noro 4273: }
1.133 noro 4274: }
1.135 noro 4275: nz = compress_array(svect,cvect,col);
4276: removecont_array(cvect,nz);
4277: expand_array(svect,cvect,nz);
1.133 noro 4278: if ( DP_Print ) {
4279: fprintf(asir_out,"-"); fflush(asir_out);
1.107 noro 4280: }
4281: return maxrs;
4282: }
4283:
1.76 noro 4284: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4285: {
1.67 noro 4286: int i,j,k,len,pos,prev;
1.66 noro 4287: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4288: IndArray ivect;
4289: unsigned char *ivc;
4290: unsigned short *ivs;
4291: unsigned int *ivi;
1.65 noro 4292: NDV redv;
1.67 noro 4293: NMV mr;
1.65 noro 4294: NODE rp;
1.76 noro 4295: int maxrs;
1.65 noro 4296:
1.76 noro 4297: maxrs = 0;
1.74 noro 4298: for ( i = 0; i < nred; i++ ) {
1.65 noro 4299: ivect = imat[i];
1.67 noro 4300: k = ivect->head; svect[k] %= m;
1.65 noro 4301: if ( c = svect[k] ) {
1.76 noro 4302: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4303: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4304: len = LEN(redv); mr = BDY(redv);
4305: svect[k] = 0; prev = k;
4306: switch ( ivect->width ) {
4307: case 1:
4308: ivc = ivect->index.c;
4309: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4310: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4311: prev = pos;
4312: DMA(c1,c,c2,up,lo);
4313: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4314: } else svect[pos] = lo;
4315: }
4316: break;
4317: case 2:
4318: ivs = ivect->index.s;
4319: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4320: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4321: prev = pos;
4322: DMA(c1,c,c2,up,lo);
4323: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4324: } else svect[pos] = lo;
4325: }
4326: break;
4327: case 4:
4328: ivi = ivect->index.i;
4329: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4330: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4331: prev = pos;
4332: DMA(c1,c,c2,up,lo);
4333: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4334: } else svect[pos] = lo;
4335: }
4336: break;
1.65 noro 4337: }
4338: }
4339: }
1.66 noro 4340: for ( i = 0; i < col; i++ )
4341: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4342: return maxrs;
1.65 noro 4343: }
4344:
1.76 noro 4345: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4346: {
4347: int i,j,k,len,pos,prev;
4348: UINT c,c1,c2,c3,up,lo,dmy;
4349: IndArray ivect;
4350: unsigned char *ivc;
4351: unsigned short *ivs;
4352: unsigned int *ivi;
4353: NDV redv;
4354: NMV mr;
4355: NODE rp;
1.76 noro 4356: int maxrs;
1.72 noro 4357:
1.76 noro 4358: maxrs = 0;
1.74 noro 4359: for ( i = 0; i < nred; i++ ) {
1.72 noro 4360: ivect = imat[i];
4361: k = ivect->head; svect[k] %= m;
4362: if ( c = svect[k] ) {
1.76 noro 4363: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4364: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4365: len = LEN(redv); mr = BDY(redv);
4366: svect[k] = 0; prev = k;
4367: switch ( ivect->width ) {
4368: case 1:
4369: ivc = ivect->index.c;
4370: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4371: pos = prev+ivc[j]; prev = pos;
4372: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4373: }
4374: break;
4375: case 2:
4376: ivs = ivect->index.s;
4377: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4378: pos = prev+ivs[j]; prev = pos;
4379: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4380: }
4381: break;
4382: case 4:
4383: ivi = ivect->index.i;
4384: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4385: pos = prev+ivi[j]; prev = pos;
4386: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4387: }
4388: break;
4389: }
4390: }
4391: }
1.76 noro 4392: return maxrs;
1.72 noro 4393: }
4394:
1.65 noro 4395: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4396: {
4397: int j,k,len;
4398: UINT *p;
4399: UINT c;
4400: NDV r;
4401: NMV mr0,mr;
4402:
4403: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4404: if ( !len ) return 0;
4405: else {
1.74 noro 4406: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4407: #if 0
1.74 noro 4408: ndv_alloc += nmv_adv*len;
1.103 noro 4409: #endif
1.65 noro 4410: mr = mr0;
4411: p = s0vect;
4412: for ( j = k = 0; j < col; j++, p += nd_wpd )
4413: if ( !rhead[j] ) {
4414: if ( c = vect[k++] ) {
4415: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4416: }
4417: }
4418: MKNDV(nd_nvar,mr0,len,r);
4419: return r;
4420: }
4421: }
4422:
1.129 noro 4423: /* for preprocessed vector */
4424:
1.113 noro 4425: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4426: {
4427: int j,k,len;
4428: UINT *p;
1.113 noro 4429: Q c;
1.107 noro 4430: NDV r;
4431: NMV mr0,mr;
4432:
4433: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4434: if ( !len ) return 0;
4435: else {
1.109 noro 4436: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4437: #if 0
4438: ndv_alloc += nmv_adv*len;
4439: #endif
4440: mr = mr0;
4441: p = s0vect;
4442: for ( j = k = 0; j < col; j++, p += nd_wpd )
4443: if ( !rhead[j] ) {
4444: if ( c = vect[k++] ) {
1.113 noro 4445: if ( DN(c) )
4446: error("afo");
4447: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4448: }
4449: }
4450: MKNDV(nd_nvar,mr0,len,r);
4451: return r;
4452: }
4453: }
4454:
1.129 noro 4455: /* for plain vector */
4456:
4457: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4458: {
4459: int j,k,len;
4460: UINT *p;
4461: Q c;
4462: NDV r;
4463: NMV mr0,mr;
4464:
4465: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4466: if ( !len ) return 0;
4467: else {
4468: mr0 = (NMV)GC_malloc(nmv_adv*len);
4469: #if 0
4470: ndv_alloc += nmv_adv*len;
4471: #endif
4472: mr = mr0;
4473: p = s0vect;
4474: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4475: if ( c = vect[k] ) {
4476: if ( DN(c) )
4477: error("afo");
4478: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4479: }
4480: MKNDV(nd_nvar,mr0,len,r);
4481: return r;
4482: }
4483: }
4484:
1.133 noro 4485: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 4486: {
4487: ND_pairs t;
4488: NODE sp0,sp;
4489: int stat;
4490: ND spol;
4491:
4492: for ( t = l; t; t = NEXT(t) ) {
1.133 noro 4493: stat = nd_sp(m,trace,t,&spol);
1.65 noro 4494: if ( !stat ) return 0;
4495: if ( spol ) {
4496: add_pbucket_symbolic(bucket,spol);
4497: }
4498: }
1.68 noro 4499: return 1;
1.65 noro 4500: }
4501:
1.133 noro 4502: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 4503: {
4504: NODE rp0,rp;
4505: NM mul,head,s0,s;
1.76 noro 4506: int index,col,i,sugar;
1.65 noro 4507: RHist h;
4508: UINT *s0v,*p;
4509: NM_ind_pair pair;
4510: ND red;
1.133 noro 4511: NDV *ps;
1.65 noro 4512:
4513: s0 = 0; rp0 = 0; col = 0;
1.133 noro 4514: ps = trace?nd_ps_trace:nd_ps;
1.65 noro 4515: while ( 1 ) {
4516: head = remove_head_pbucket_symbolic(bucket);
4517: if ( !head ) break;
4518: if ( !s0 ) s0 = head;
4519: else NEXT(s) = head;
4520: s = head;
4521: index = ndl_find_reducer(DL(head));
4522: if ( index >= 0 ) {
4523: h = nd_psh[index];
4524: NEWNM(mul);
4525: ndl_sub(DL(head),DL(h),DL(mul));
4526: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.133 noro 4527: sugar = TD(DL(mul))+SG(ps[index]);
1.76 noro 4528: MKNM_ind_pair(pair,mul,index,sugar);
1.133 noro 4529: red = ndv_mul_nm_symbolic(mul,ps[index]);
1.65 noro 4530: add_pbucket_symbolic(bucket,nd_remove_head(red));
4531: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4532: }
4533: col++;
4534: }
1.72 noro 4535: if ( rp0 ) NEXT(rp) = 0;
4536: NEXT(s) = 0;
1.65 noro 4537: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4538: for ( i = 0, p = s0v, s = s0; i < col;
4539: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4540: *s0vect = s0v;
4541: *r = rp0;
4542: return col;
4543: }
4544:
1.69 noro 4545: NODE nd_f4(int m)
4546: {
4547: int i,nh,stat,index;
4548: NODE r,g;
4549: ND_pairs d,l,t;
4550: ND spol,red;
4551: NDV nf,redv;
4552: NM s0,s;
1.84 noro 4553: NODE rp0,srp0,nflist;
1.69 noro 4554: int nsp,nred,col,rank,len,k,j,a;
4555: UINT c;
1.74 noro 4556: UINT **spmat;
1.69 noro 4557: UINT *s0vect,*svect,*p,*v;
4558: int *colstat;
4559: IndArray *imat;
4560: int *rhead;
4561: int spcol,sprow;
4562: int sugar;
4563: PGeoBucket bucket;
4564: struct oEGT eg0,eg1,eg_f4;
4565:
1.103 noro 4566: #if 0
1.74 noro 4567: ndv_alloc = 0;
1.103 noro 4568: #endif
1.69 noro 4569: g = 0; d = 0;
4570: for ( i = 0; i < nd_psn; i++ ) {
4571: d = update_pairs(d,g,i);
4572: g = update_base(g,i);
4573: }
4574: while ( d ) {
4575: get_eg(&eg0);
4576: l = nd_minsugarp(d,&d);
4577: sugar = SG(l);
4578: bucket = create_pbucket();
1.133 noro 4579: stat = nd_sp_f4(m,0,l,bucket);
1.69 noro 4580: if ( !stat ) {
4581: for ( t = l; NEXT(t); t = NEXT(t) );
4582: NEXT(t) = d; d = l;
1.103 noro 4583: d = nd_reconstruct(0,d);
1.69 noro 4584: continue;
4585: }
1.86 noro 4586: if ( bucket->m < 0 ) continue;
1.133 noro 4587: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
1.69 noro 4588: if ( !col ) {
4589: for ( t = l; NEXT(t); t = NEXT(t) );
4590: NEXT(t) = d; d = l;
1.103 noro 4591: d = nd_reconstruct(0,d);
1.69 noro 4592: continue;
4593: }
4594: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4595: if ( DP_Print )
4596: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4597: sugar,eg_f4.exectime+eg_f4.gctime);
4598: if ( 1 )
1.133 noro 4599: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
1.74 noro 4600: else
1.133 noro 4601: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
1.69 noro 4602: /* adding new bases */
1.74 noro 4603: for ( r = nflist; r; r = NEXT(r) ) {
4604: nf = (NDV)BDY(r);
1.69 noro 4605: ndv_removecont(m,nf);
1.129 noro 4606: if ( !m && nd_nalg ) {
4607: ND nf1;
4608:
4609: nf1 = ndvtond(m,nf);
4610: nd_monic(0,&nf1);
4611: nd_removecont(m,nf1);
4612: nf = ndtondv(m,nf1);
4613: }
1.77 noro 4614: nh = ndv_newps(m,nf,0);
1.69 noro 4615: d = update_pairs(d,g,nh);
4616: g = update_base(g,nh);
4617: }
4618: }
4619: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4620: #if 0
1.74 noro 4621: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4622: #endif
1.69 noro 4623: return g;
4624: }
1.74 noro 4625:
1.133 noro 4626: NODE nd_f4_trace(int m)
4627: {
4628: int i,nh,stat,index;
4629: NODE r,g;
4630: ND_pairs d,l,l0,t;
4631: ND spol,red;
4632: NDV nf,redv,nfqv,nfv;
4633: NM s0,s;
4634: NODE rp0,srp0,nflist;
4635: int nsp,nred,col,rank,len,k,j,a;
4636: UINT c;
4637: UINT **spmat;
4638: UINT *s0vect,*svect,*p,*v;
4639: int *colstat;
4640: IndArray *imat;
4641: int *rhead;
4642: int spcol,sprow;
4643: int sugar;
4644: PGeoBucket bucket;
4645: struct oEGT eg0,eg1,eg_f4;
4646:
4647: g = 0; d = 0;
4648: for ( i = 0; i < nd_psn; i++ ) {
4649: d = update_pairs(d,g,i);
4650: g = update_base(g,i);
4651: }
4652: while ( d ) {
4653: get_eg(&eg0);
4654: l = nd_minsugarp(d,&d);
4655: sugar = SG(l);
4656: bucket = create_pbucket();
4657: stat = nd_sp_f4(m,0,l,bucket);
4658: if ( !stat ) {
4659: for ( t = l; NEXT(t); t = NEXT(t) );
4660: NEXT(t) = d; d = l;
4661: d = nd_reconstruct(1,d);
4662: continue;
4663: }
4664: if ( bucket->m < 0 ) continue;
4665: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
4666: if ( !col ) {
4667: for ( t = l; NEXT(t); t = NEXT(t) );
4668: NEXT(t) = d; d = l;
4669: d = nd_reconstruct(1,d);
4670: continue;
4671: }
4672: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4673: if ( DP_Print )
4674: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4675: sugar,eg_f4.exectime+eg_f4.gctime);
4676: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
4677: if ( !l0 ) continue;
4678: l = l0;
4679:
4680: /* over Q */
4681: bucket = create_pbucket();
4682: stat = nd_sp_f4(0,1,l,bucket);
4683: if ( !stat ) {
4684: for ( t = l; NEXT(t); t = NEXT(t) );
4685: NEXT(t) = d; d = l;
4686: d = nd_reconstruct(1,d);
4687: continue;
4688: }
4689: if ( bucket->m < 0 ) continue;
4690: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
4691: if ( !col ) {
4692: for ( t = l; NEXT(t); t = NEXT(t) );
4693: NEXT(t) = d; d = l;
4694: d = nd_reconstruct(1,d);
4695: continue;
4696: }
4697: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
4698: /* adding new bases */
4699: for ( r = nflist; r; r = NEXT(r) ) {
4700: nfqv = (NDV)BDY(r);
4701: ndv_removecont(0,nfqv);
4702: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
4703: if ( nd_nalg ) {
4704: ND nf1;
4705:
4706: nf1 = ndvtond(m,nfqv);
4707: nd_monic(0,&nf1);
4708: nd_removecont(0,nf1);
4709: nfqv = ndtondv(0,nf1); nd_free(nf1);
4710: }
4711: nfv = ndv_dup(0,nfqv);
4712: ndv_mod(m,nfv);
4713: ndv_removecont(m,nfv);
4714: nh = ndv_newps(0,nfv,nfqv);
4715: d = update_pairs(d,g,nh);
4716: g = update_base(g,nh);
4717: }
4718: }
4719: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
4720: #if 0
4721: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
4722: #endif
4723: return g;
4724: }
4725:
4726: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 4727: {
1.67 noro 4728: IndArray *imat;
1.106 noro 4729: int nsp,nred,i;
1.65 noro 4730: int *rhead;
1.106 noro 4731: NODE r0,rp;
1.74 noro 4732: ND_pairs sp;
4733: NM_ind_pair *rvect;
1.63 noro 4734:
1.74 noro 4735: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4736: nred = length(rp0);
1.74 noro 4737: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4738: rhead = (int *)ALLOCA(col*sizeof(int));
4739: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4740:
1.74 noro 4741: /* construction of index arrays */
4742: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4743: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4744: rvect[i] = (NM_ind_pair)BDY(rp);
4745: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4746: rhead[imat[i]->head] = 1;
4747: }
1.107 noro 4748: if ( m )
1.133 noro 4749: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
1.107 noro 4750: else
1.133 noro 4751: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4752: return r0;
4753: }
1.74 noro 4754:
1.106 noro 4755: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 4756: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 4757: {
4758: int spcol,sprow,a;
4759: int i,j,k,l,rank;
4760: NODE r0,r;
4761: ND_pairs sp;
4762: ND spol;
4763: int **spmat;
4764: UINT *svect,*v;
4765: int *colstat;
4766: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4767: int maxrs;
4768: int *spsugar;
1.133 noro 4769: ND_pairs *spactive;
1.106 noro 4770:
4771: spcol = col-nred;
4772: get_eg(&eg0);
1.74 noro 4773: /* elimination (1st step) */
4774: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4775: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4776: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.133 noro 4777: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
1.74 noro 4778: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4779: nd_sp(m,0,sp,&spol);
1.75 noro 4780: if ( !spol ) continue;
1.74 noro 4781: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4782: if ( m == -1 )
4783: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4784: else
4785: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4786: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4787: if ( i < col ) {
4788: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4789: for ( j = k = 0; j < col; j++ )
4790: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4791: spsugar[sprow] = MAX(maxrs,SG(spol));
1.133 noro 4792: if ( nz )
4793: spactive[sprow] = sp;
1.74 noro 4794: sprow++;
4795: }
1.76 noro 4796: nd_free(spol);
1.74 noro 4797: }
1.80 noro 4798: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4799: if ( DP_Print ) {
4800: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4801: fflush(asir_out);
4802: }
1.74 noro 4803: /* free index arrays */
4804: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4805:
4806: /* elimination (2nd step) */
4807: colstat = (int *)ALLOCA(spcol*sizeof(int));
4808: if ( m == -1 )
1.76 noro 4809: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4810: else
1.133 noro 4811: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
1.74 noro 4812: r0 = 0;
4813: for ( i = 0; i < rank; i++ ) {
4814: NEXTNODE(r0,r); BDY(r) =
4815: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4816: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4817: GC_free(spmat[i]);
4818: }
1.109 noro 4819: if ( r0 ) NEXT(r) = 0;
1.133 noro 4820:
1.74 noro 4821: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4822: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4823: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4824: if ( DP_Print ) {
1.80 noro 4825: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4826: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4827: nsp,nred,sprow,spcol,rank);
4828: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4829: }
1.133 noro 4830: if ( nz ) {
4831: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
4832: if ( rank > 0 ) {
4833: NEXT(spactive[rank-1]) = 0;
4834: *nz = spactive[0];
4835: } else
4836: *nz = 0;
4837: }
1.74 noro 4838: return r0;
4839: }
4840:
1.133 noro 4841: #if 1
4842: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 4843: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4844: {
4845: int spcol,sprow,a;
4846: int i,j,k,l,rank;
4847: NODE r0,r;
4848: ND_pairs sp;
4849: ND spol;
1.113 noro 4850: Q **spmat;
4851: Q *svect,*v;
1.107 noro 4852: int *colstat;
4853: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4854: int maxrs;
4855: int *spsugar;
1.137 noro 4856: pointer *w;
1.107 noro 4857:
4858: spcol = col-nred;
4859: get_eg(&eg0);
4860: /* elimination (1st step) */
1.113 noro 4861: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4862: svect = (Q *)ALLOCA(col*sizeof(Q));
4863: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4864: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
1.133 noro 4865: nd_sp(0,trace,sp,&spol);
1.107 noro 4866: if ( !spol ) continue;
1.113 noro 4867: nd_to_vect_q(s0vect,col,spol,svect);
1.133 noro 4868: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
1.107 noro 4869: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4870: if ( i < col ) {
1.113 noro 4871: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4872: for ( j = k = 0; j < col; j++ )
4873: if ( !rhead[j] ) v[k++] = svect[j];
4874: spsugar[sprow] = MAX(maxrs,SG(spol));
4875: sprow++;
4876: }
1.109 noro 4877: /* nd_free(spol); */
1.107 noro 4878: }
4879: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4880: if ( DP_Print ) {
4881: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4882: fflush(asir_out);
4883: }
4884: /* free index arrays */
1.109 noro 4885: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4886:
4887: /* elimination (2nd step) */
4888: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4889: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.137 noro 4890: w = (pointer *)ALLOCA(rank*sizeof(pointer));
4891: for ( i = 0; i < rank; i++ ) {
4892: w[rank-i-1] = (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
4893: SG((NDV)w[rank-i-1]) = spsugar[i];
4894: /* GC_free(spmat[i]); */
4895: }
1.138 noro 4896: #if 0
1.137 noro 4897: qsort(w,rank,sizeof(NDV),
4898: (int (*)(const void *,const void *))ndv_compare);
4899: #endif
1.107 noro 4900: r0 = 0;
4901: for ( i = 0; i < rank; i++ ) {
1.137 noro 4902: NEXTNODE(r0,r); BDY(r) = w[i];
1.107 noro 4903: }
1.137 noro 4904: if ( r0 ) NEXT(r) = 0;
1.135 noro 4905:
1.109 noro 4906: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4907: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4908: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
4909: if ( DP_Print ) {
4910: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
4911: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4912: nsp,nred,sprow,spcol,rank);
4913: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
4914: }
4915: return r0;
4916: }
1.129 noro 4917: #else
4918: void printm(Q **mat,int row,int col)
4919: {
4920: int i,j;
4921: printf("[");
4922: for ( i = 0; i < row; i++ ) {
4923: for ( j = 0; j < col; j++ ) {
4924: printexpr(CO,mat[i][j]); printf(" ");
4925: }
4926: printf("]\n");
4927: }
4928: }
4929:
4930: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
4931: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4932: {
4933: int row,a;
4934: int i,j,rank;
4935: NODE r0,r;
4936: ND_pairs sp;
4937: ND spol;
4938: Q **mat;
4939: int *colstat;
4940: int *sugar;
4941:
4942: row = nsp+nred;
4943: /* make the matrix */
4944: mat = (Q **)ALLOCA(row*sizeof(Q *));
4945: sugar = (int *)ALLOCA(row*sizeof(int));
4946: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4947: nd_sp(0,0,sp,&spol);
4948: if ( !spol ) continue;
4949: mat[row] = (Q *)MALLOC(col*sizeof(Q));
4950: nd_to_vect_q(s0vect,col,spol,mat[row]);
4951: sugar[row] = SG(spol);
4952: row++;
4953: }
4954: for ( i = 0; i < nred; i++, row++ ) {
4955: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
4956: sugar[row] = rvect[i]->sugar;
4957: }
4958: /* elimination */
4959: colstat = (int *)ALLOCA(col*sizeof(int));
4960: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
4961: r0 = 0;
4962: for ( i = 0; i < rank; i++ ) {
4963: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
4964: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
4965: if ( rhead[j] ) continue;
4966: NEXTNODE(r0,r); BDY(r) =
4967: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
4968: SG((NDV)BDY(r)) = sugar[i];
4969: }
4970: if ( r0 ) NEXT(r) = 0;
4971: printf("\n");
4972: return r0;
4973: }
4974: #endif
1.107 noro 4975:
1.74 noro 4976: FILE *nd_write,*nd_read;
4977:
4978: void nd_send_int(int a) {
4979: write_int(nd_write,&a);
4980: }
4981:
4982: void nd_send_intarray(int *p,int len) {
4983: write_intarray(nd_write,p,len);
4984: }
4985:
4986: int nd_recv_int() {
4987: int a;
4988:
4989: read_int(nd_read,&a);
4990: return a;
4991: }
4992:
4993: void nd_recv_intarray(int *p,int len) {
4994: read_intarray(nd_read,p,len);
4995: }
4996:
4997: void nd_send_ndv(NDV p) {
4998: int len,i;
4999: NMV m;
5000:
5001: if ( !p ) nd_send_int(0);
5002: else {
5003: len = LEN(p);
5004: nd_send_int(len);
5005: m = BDY(p);
5006: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5007: nd_send_int(CM(m));
5008: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 5009: }
1.74 noro 5010: }
5011: }
5012:
5013: void nd_send_nd(ND p) {
5014: int len,i;
5015: NM m;
5016:
5017: if ( !p ) nd_send_int(0);
5018: else {
5019: len = LEN(p);
5020: nd_send_int(len);
5021: m = BDY(p);
5022: for ( i = 0; i < len; i++, m = NEXT(m) ) {
5023: nd_send_int(CM(m));
5024: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 5025: }
1.74 noro 5026: }
5027: }
1.65 noro 5028:
1.74 noro 5029: NDV nd_recv_ndv()
5030: {
5031: int len,i;
5032: NMV m,m0;
5033: NDV r;
1.65 noro 5034:
1.74 noro 5035: len = nd_recv_int();
5036: if ( !len ) return 0;
5037: else {
5038: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5039: #if 0
1.74 noro 5040: ndv_alloc += len*nmv_adv;
1.103 noro 5041: #endif
1.74 noro 5042: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5043: CM(m) = nd_recv_int();
5044: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 5045: }
1.74 noro 5046: MKNDV(nd_nvar,m0,len,r);
5047: return r;
5048: }
5049: }
1.65 noro 5050:
1.74 noro 5051: int ox_exec_f4_red(Q proc)
5052: {
5053: Obj obj;
5054: STRING fname;
5055: NODE arg;
5056: int s;
5057: extern int ox_need_conv,ox_file_io;
5058:
5059: MKSTR(fname,"nd_exec_f4_red");
5060: arg = mknode(2,proc,fname);
5061: Pox_cmo_rpc(arg,&obj);
5062: s = get_ox_server_id(QTOS(proc));
5063: nd_write = iofp[s].out;
5064: nd_read = iofp[s].in;
5065: ox_need_conv = ox_file_io = 0;
5066: return s;
5067: }
5068:
1.133 noro 5069: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5070: {
5071: int nsp,nred;
5072: int i,rank,s;
5073: NODE rp,r0,r;
5074: ND_pairs sp;
5075: NM_ind_pair pair;
5076: NMV nmv;
5077: NM nm;
5078: NDV nf;
5079: Obj proc,dmy;
5080:
5081: ox_launch_main(0,0,&proc);
5082: s = ox_exec_f4_red((Q)proc);
5083:
5084: nd_send_int(m);
5085: nd_send_int(nd_nvar);
5086: nd_send_int(nd_bpe);
5087: nd_send_int(nd_wpd);
5088: nd_send_int(nmv_adv);
5089:
1.89 noro 5090: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 5091:
5092: nd_send_int(nd_psn);
5093: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5094:
5095: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5096: nd_send_int(nsp);
5097: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5098: nd_send_int(sp->i1); nd_send_int(sp->i2);
5099: }
5100:
5101: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5102:
5103: nred = length(rp0); nd_send_int(nred);
5104: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5105: pair = (NM_ind_pair)BDY(rp);
5106: nd_send_int(pair->index);
5107: nd_send_intarray(pair->mul->dl,nd_wpd);
5108: }
5109: fflush(nd_write);
5110: rank = nd_recv_int();
5111: fprintf(asir_out,"rank=%d\n",rank);
5112: r0 = 0;
5113: for ( i = 0; i < rank; i++ ) {
5114: nf = nd_recv_ndv();
5115: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5116: }
5117: Pox_shutdown(mknode(1,proc),&dmy);
5118: return r0;
5119: }
5120:
5121: /* server side */
5122:
5123: void nd_exec_f4_red_dist()
5124: {
5125: int m,i,nsp,col,s0size,nred,spcol,j,k;
5126: NM_ind_pair *rp0;
5127: NDV nf;
5128: UINT *s0vect;
5129: IndArray *imat;
5130: int *rhead;
5131: int **spmat;
5132: UINT *svect,*v;
5133: ND_pairs *sp0;
5134: int *colstat;
5135: int a,sprow,rank;
1.89 noro 5136: struct order_spec *ord;
1.74 noro 5137: Obj ordspec;
5138: ND spol;
1.76 noro 5139: int maxrs;
5140: int *spsugar;
1.74 noro 5141:
5142: nd_read = iofp[0].in;
5143: nd_write = iofp[0].out;
5144: m = nd_recv_int();
5145: nd_nvar = nd_recv_int();
5146: nd_bpe = nd_recv_int();
5147: nd_wpd = nd_recv_int();
5148: nmv_adv = nd_recv_int();
5149:
5150: loadobj(nd_read,&ordspec);
1.89 noro 5151: create_order_spec(0,ordspec,&ord);
5152: nd_init_ord(ord);
1.74 noro 5153: nd_setup_parameters(nd_nvar,0);
5154:
5155: nd_psn = nd_recv_int();
5156: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5157: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5158: for ( i = 0; i < nd_psn; i++ ) {
5159: nd_ps[i] = nd_recv_ndv();
5160: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5161: }
5162:
5163: nsp = nd_recv_int();
5164: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5165: for ( i = 0; i < nsp; i++ ) {
5166: NEWND_pairs(sp0[i]);
5167: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5168: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5169: }
5170:
5171: col = nd_recv_int();
5172: s0size = col*nd_wpd;
5173: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5174: nd_recv_intarray(s0vect,s0size);
5175:
5176: nred = nd_recv_int();
5177: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5178: for ( i = 0; i < nred; i++ ) {
5179: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5180: rp0[i]->index = nd_recv_int();
5181: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5182: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5183: }
5184:
5185: spcol = col-nred;
5186: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5187: rhead = (int *)MALLOC(col*sizeof(int));
5188: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5189:
5190: /* construction of index arrays */
5191: for ( i = 0; i < nred; i++ ) {
5192: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5193: rhead[imat[i]->head] = 1;
5194: }
5195:
5196: /* elimination (1st step) */
5197: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5198: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5199: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5200: for ( a = sprow = 0; a < nsp; a++ ) {
5201: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5202: if ( !spol ) continue;
1.74 noro 5203: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5204: if ( m == -1 )
5205: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5206: else
5207: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5208: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5209: if ( i < col ) {
5210: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5211: for ( j = k = 0; j < col; j++ )
5212: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5213: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5214: sprow++;
5215: }
1.76 noro 5216: nd_free(spol);
1.74 noro 5217: }
5218: /* elimination (2nd step) */
5219: colstat = (int *)ALLOCA(spcol*sizeof(int));
5220: if ( m == -1 )
1.76 noro 5221: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5222: else
1.133 noro 5223: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
1.74 noro 5224: nd_send_int(rank);
5225: for ( i = 0; i < rank; i++ ) {
5226: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5227: nd_send_ndv(nf);
1.63 noro 5228: }
1.74 noro 5229: fflush(nd_write);
1.107 noro 5230: }
5231:
1.113 noro 5232: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5233: {
1.109 noro 5234: int mod,i,j,t,c,rank,rank0,inv;
5235: int *ci,*ri;
1.113 noro 5236: Q dn;
1.109 noro 5237: MAT m,nm;
1.108 noro 5238: int **wmat;
5239:
5240: /* XXX */
5241: mod = 99999989;
5242: wmat = (int **)ALLOCA(row*sizeof(int *));
5243: for ( i = 0; i < row; i++ ) {
5244: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5245: for ( j = 0; j < col; j++ ) {
1.113 noro 5246: if ( mat0[i][j] ) {
5247: t = rem(NM(mat0[i][j]),mod);
5248: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5249: wmat[i][j] = t;
5250: } else
1.108 noro 5251: wmat[i][j] = 0;
5252: }
5253: }
1.133 noro 5254: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
1.109 noro 5255: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5256: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5257: if ( rank != rank0 )
5258: error("afo");
5259: for ( i = 0; i < row; i++ )
5260: for ( j = 0; j < col; j++ )
5261: mat0[i][j] = 0;
5262: c = col-rank;
5263: for ( i = 0; i < rank; i++ ) {
1.113 noro 5264: mat0[i][ri[i]] = dn;
1.109 noro 5265: for ( j = 0; j < c; j++ )
1.113 noro 5266: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5267: }
1.113 noro 5268: inv = invm(rem(NM(dn),mod),mod);
5269: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5270: for ( i = 0; i < row; i++ )
5271: for ( j = 0; j < col; j++ ) {
1.113 noro 5272: if ( mat0[i][j] ) {
5273: t = rem(NM(mat0[i][j]),mod);
5274: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5275: } else
1.109 noro 5276: t = 0;
5277: c = dmar(t,inv,0,mod);
5278: if ( wmat[i][j] != c )
5279: error("afo");
5280: }
5281: return rank;
1.76 noro 5282: }
5283:
1.133 noro 5284: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5285: {
5286: int i,j,k,l,inv,a,rank,s;
5287: unsigned int *t,*pivot,*pk;
5288: unsigned int **mat;
1.133 noro 5289: ND_pairs pair;
1.76 noro 5290:
5291: mat = (unsigned int **)mat0;
5292: for ( rank = 0, j = 0; j < col; j++ ) {
5293: for ( i = rank; i < row; i++ )
5294: mat[i][j] %= md;
5295: for ( i = rank; i < row; i++ )
5296: if ( mat[i][j] )
5297: break;
5298: if ( i == row ) {
5299: colstat[j] = 0;
5300: continue;
5301: } else
5302: colstat[j] = 1;
5303: if ( i != rank ) {
5304: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5305: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
1.133 noro 5306: if ( spactive ) {
5307: pair = spactive[i]; spactive[i] = spactive[rank];
5308: spactive[rank] = pair;
5309: }
1.76 noro 5310: }
5311: pivot = mat[rank];
5312: s = sugar[rank];
5313: inv = invm(pivot[j],md);
5314: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5315: if ( *pk ) {
5316: if ( *pk >= (unsigned int)md )
5317: *pk %= md;
5318: DMAR(*pk,inv,0,md,*pk)
5319: }
5320: for ( i = rank+1; i < row; i++ ) {
5321: t = mat[i];
5322: if ( a = t[j] ) {
5323: sugar[i] = MAX(sugar[i],s);
5324: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5325: }
5326: }
5327: rank++;
5328: }
5329: for ( j = col-1, l = rank-1; j >= 0; j-- )
5330: if ( colstat[j] ) {
5331: pivot = mat[l];
5332: s = sugar[l];
5333: for ( i = 0; i < l; i++ ) {
5334: t = mat[i];
5335: t[j] %= md;
5336: if ( a = t[j] ) {
5337: sugar[i] = MAX(sugar[i],s);
5338: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5339: }
5340: }
5341: l--;
5342: }
5343: for ( j = 0, l = 0; l < rank; j++ )
5344: if ( colstat[j] ) {
5345: t = mat[l];
5346: for ( k = j; k < col; k++ )
5347: if ( t[k] >= (unsigned int)md )
5348: t[k] %= md;
5349: l++;
5350: }
5351: return rank;
5352: }
5353:
5354: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5355: {
5356: int i,j,k,l,inv,a,rank,s;
5357: unsigned int *t,*pivot,*pk;
5358: unsigned int **mat;
5359:
5360: mat = (unsigned int **)mat0;
5361: for ( rank = 0, j = 0; j < col; j++ ) {
5362: for ( i = rank; i < row; i++ )
5363: if ( mat[i][j] )
5364: break;
5365: if ( i == row ) {
5366: colstat[j] = 0;
5367: continue;
5368: } else
5369: colstat[j] = 1;
5370: if ( i != rank ) {
5371: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5372: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5373: }
5374: pivot = mat[rank];
5375: s = sugar[rank];
5376: inv = _invsf(pivot[j]);
5377: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5378: if ( *pk )
5379: *pk = _mulsf(*pk,inv);
5380: for ( i = rank+1; i < row; i++ ) {
5381: t = mat[i];
5382: if ( a = t[j] ) {
5383: sugar[i] = MAX(sugar[i],s);
5384: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5385: }
5386: }
5387: rank++;
5388: }
5389: for ( j = col-1, l = rank-1; j >= 0; j-- )
5390: if ( colstat[j] ) {
5391: pivot = mat[l];
5392: s = sugar[l];
5393: for ( i = 0; i < l; i++ ) {
5394: t = mat[i];
5395: if ( a = t[j] ) {
5396: sugar[i] = MAX(sugar[i],s);
5397: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5398: }
5399: }
5400: l--;
5401: }
5402: return rank;
1.77 noro 5403: }
5404:
5405: int ndv_ishomo(NDV p)
5406: {
5407: NMV m;
5408: int len,h;
5409:
5410: if ( !p ) return 1;
5411: len = LEN(p);
5412: m = BDY(p);
5413: h = TD(DL(m));
5414: NMV_ADV(m);
5415: for ( len--; len; len--, NMV_ADV(m) )
5416: if ( TD(DL(m)) != h ) return 0;
5417: return 1;
5418: }
5419:
5420: void ndv_save(NDV p,int index)
5421: {
5422: FILE *s;
5423: char name[BUFSIZ];
5424: short id;
5425: int nv,sugar,len,n,i,td,e,j;
5426: NMV m;
5427: unsigned int *dl;
5428:
5429: sprintf(name,"%s/%d",Demand,index);
5430: s = fopen(name,"w");
5431: savevl(s,0);
5432: if ( !p ) {
5433: saveobj(s,0);
5434: return;
5435: }
5436: id = O_DP;
5437: nv = NV(p);
5438: sugar = SG(p);
5439: len = LEN(p);
5440: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5441: write_int(s,&len);
5442:
5443: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5444: saveobj(s,(Obj)CQ(m));
1.77 noro 5445: dl = DL(m);
5446: td = TD(dl);
5447: write_int(s,&td);
5448: for ( j = 0; j < nv; j++ ) {
5449: e = GET_EXP(dl,j);
5450: write_int(s,&e);
5451: }
5452: }
5453: fclose(s);
5454: }
5455:
5456: NDV ndv_load(int index)
5457: {
5458: FILE *s;
5459: char name[BUFSIZ];
5460: short id;
5461: int nv,sugar,len,n,i,td,e,j;
5462: NDV d;
5463: NMV m0,m;
5464: unsigned int *dl;
5465: Obj obj;
5466:
5467: sprintf(name,"%s/%d",Demand,index);
5468: s = fopen(name,"r");
5469: if ( !s ) return 0;
5470:
5471: skipvl(s);
5472: read_short(s,&id);
5473: if ( !id ) return 0;
5474: read_int(s,&nv);
5475: read_int(s,&sugar);
5476: read_int(s,&len);
5477:
5478: m0 = m = MALLOC(len*nmv_adv);
5479: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5480: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5481: dl = DL(m);
5482: ndl_zero(dl);
5483: read_int(s,&td); TD(dl) = td;
5484: for ( j = 0; j < nv; j++ ) {
5485: read_int(s,&e);
5486: PUT_EXP(dl,j,e);
5487: }
5488: if ( nd_blockmask ) ndl_weight_mask(dl);
5489: }
5490: fclose(s);
5491: MKNDV(nv,m0,len,d);
5492: SG(d) = sugar;
5493: return d;
1.99 noro 5494: }
5495:
1.102 noro 5496: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5497: {
5498: VL fv,tv;
1.102 noro 5499: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5500: pointer **m;
1.113 noro 5501: Q mone;
1.125 noro 5502: P **w;
1.126 noro 5503: P mp,r;
1.102 noro 5504: NDV **dm;
5505: NDV *t,*mi,*mj;
5506: NDV d,s,mij,mjj;
5507: ND u;
5508: NMV nmv;
1.114 noro 5509: UINT *bound;
1.99 noro 5510: PGeoBucket bucket;
5511: struct order_spec *ord;
1.126 noro 5512: Q dq,dt,ds;
5513: N gn,qn,dn0,nm,dn;
1.99 noro 5514:
5515: create_order_spec(0,0,&ord);
5516: nd_init_ord(ord);
5517: get_vars((Obj)f,&fv);
5518: if ( f->row != f->col )
5519: error("nd_det : non-square matrix");
5520: n = f->row;
1.125 noro 5521: m = f->body;
1.99 noro 5522: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5523:
5524: if ( !nvar ) {
5525: if ( !mod )
5526: detp(CO,(P **)m,n,rp);
5527: else {
5528: w = (P **)almat_pointer(n,n);
5529: for ( i = 0; i < n; i++ )
5530: for ( j = 0; j < n; j++ )
5531: ptomp(mod,(P)m[i][j],&w[i][j]);
5532: detmp(CO,mod,w,n,&mp);
5533: mptop(mp,rp);
5534: }
5535: return;
5536: }
1.126 noro 5537:
5538: if ( !mod ) {
5539: w = (P **)almat_pointer(n,n);
5540: dq = ONE;
5541: for ( i = 0; i < n; i++ ) {
5542: dn0 = ONEN;
5543: for ( j = 0; j < n; j++ ) {
5544: if ( !m[i][j] ) continue;
5545: lgp(m[i][j],&nm,&dn);
5546: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5547: }
5548: if ( !UNIN(dn0) ) {
5549: NTOQ(dn0,1,ds);
5550: for ( j = 0; j < n; j++ )
5551: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5552: mulq(dq,ds,&dt); dq = dt;
5553: } else
5554: for ( j = 0; j < n; j++ )
5555: w[i][j] = (P)m[i][j];
5556: }
5557: m = (pointer **)w;
5558: }
1.125 noro 5559:
1.99 noro 5560: for ( i = 0, max = 0; i < n; i++ )
5561: for ( j = 0; j < n; j++ )
5562: for ( tv = fv; tv; tv = NEXT(tv) ) {
5563: e = getdeg(tv->v,(P)m[i][j]);
5564: max = MAX(e,max);
5565: }
1.114 noro 5566: nd_setup_parameters(nvar,max);
1.102 noro 5567: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5568: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5569: for ( j = 0; j < n; j++ ) {
5570: dm[i][j] = ptondv(CO,fv,m[i][j]);
5571: if ( mod ) ndv_mod(mod,dm[i][j]);
5572: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5573: }
5574: d = ptondv(CO,fv,(P)ONE);
5575: if ( mod ) ndv_mod(mod,d);
1.113 noro 5576: chsgnq(ONE,&mone);
1.99 noro 5577: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5578: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5579: for ( i = j; i < n && !dm[i][j]; i++ );
5580: if ( i == n ) {
5581: *rp = 0;
5582: return;
5583: }
1.102 noro 5584: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5585: for ( k = j; k < n; k++ )
5586: for ( l = j; l < n; l++ )
1.102 noro 5587: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5588: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5589: }
5590: if ( k0 != j ) {
5591: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5592: sgn = -sgn;
5593: }
5594: if ( l0 != j ) {
5595: for ( k = j; k < n; k++ ) {
5596: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5597: }
5598: sgn = -sgn;
5599: }
1.114 noro 5600: bound = nd_det_compute_bound(dm,n,j);
5601: if ( ndl_check_bound(bound,bound) )
5602: nd_det_reconstruct(dm,n,j,d);
5603:
1.99 noro 5604: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5605: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5606: mi = dm[i]; mij = mi[j];
1.102 noro 5607: if ( mod )
5608: ndv_mul_c(mod,mij,mod-1);
5609: else
1.113 noro 5610: ndv_mul_c_q(mij,mone);
1.99 noro 5611: for ( k = j+1; k < n; k++ ) {
1.114 noro 5612: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5613: bucket = create_pbucket();
1.104 noro 5614: if ( mi[k] ) {
1.102 noro 5615: nmv = BDY(mjj); len = LEN(mjj);
5616: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5617: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5618: add_pbucket(mod,bucket,u);
1.99 noro 5619: }
1.104 noro 5620: }
1.99 noro 5621: if ( mj[k] && mij ) {
1.102 noro 5622: nmv = BDY(mij); len = LEN(mij);
5623: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5624: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5625: add_pbucket(mod,bucket,u);
1.99 noro 5626: }
5627: }
1.104 noro 5628: u = nd_quo(mod,bucket,d);
1.102 noro 5629: mi[k] = ndtondv(mod,u);
1.99 noro 5630: }
1.114 noro 5631: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5632: }
5633: d = mjj;
5634: }
1.125 noro 5635: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5636: if ( sgn < 0 )
1.102 noro 5637: if ( mod )
5638: ndv_mul_c(mod,d,mod-1);
5639: else
1.113 noro 5640: ndv_mul_c_q(d,mone);
1.126 noro 5641: r = ndvtop(mod,CO,fv,d);
5642: if ( !mod && !UNIQ(dq) )
5643: divsp(CO,r,(P)dq,rp);
5644: else
5645: *rp = r;
1.99 noro 5646: }
5647:
1.102 noro 5648: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5649: {
5650: NM mr,mr0;
1.102 noro 5651: NM tnm;
5652: NMV m;
1.99 noro 5653: UINT *d0,*dt,*dm;
5654: int c,n,td,i,c1,c2,len;
1.113 noro 5655: Q q;
1.99 noro 5656: ND r;
5657:
5658: if ( !p ) return 0;
5659: else {
1.102 noro 5660: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5661: d0 = DL(m0);
1.102 noro 5662: td = TD(d);
1.99 noro 5663: mr0 = 0;
5664: NEWNM(tnm);
1.102 noro 5665: if ( mod ) {
5666: c = CM(m0);
5667: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5668: ndl_add(DL(m),d0,DL(tnm));
5669: if ( ndl_reducible(DL(tnm),d) ) {
5670: NEXTNM(mr0,mr);
5671: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5672: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5673: }
5674: }
5675: } else {
1.113 noro 5676: q = CQ(m0);
1.102 noro 5677: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5678: ndl_add(DL(m),d0,DL(tnm));
5679: if ( ndl_reducible(DL(tnm),d) ) {
5680: NEXTNM(mr0,mr);
1.113 noro 5681: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5682: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5683: }
1.99 noro 5684: }
5685: }
5686: if ( !mr0 )
5687: return 0;
5688: else {
5689: NEXT(mr) = 0;
1.105 noro 5690: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5691: MKND(NV(p),mr0,len,r);
5692: SG(r) = SG(p) + TD(d0);
5693: return r;
5694: }
5695: }
1.114 noro 5696: }
5697:
5698: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5699: {
5700: int i,obpe,oadv,h,k,l;
5701: static NM prev_nm_free_list;
5702: EPOS oepos;
5703:
5704: obpe = nd_bpe;
5705: oadv = nmv_adv;
5706: oepos = nd_epos;
5707: if ( obpe < 2 ) nd_bpe = 2;
5708: else if ( obpe < 3 ) nd_bpe = 3;
5709: else if ( obpe < 4 ) nd_bpe = 4;
5710: else if ( obpe < 5 ) nd_bpe = 5;
5711: else if ( obpe < 6 ) nd_bpe = 6;
5712: else if ( obpe < 8 ) nd_bpe = 8;
5713: else if ( obpe < 10 ) nd_bpe = 10;
5714: else if ( obpe < 16 ) nd_bpe = 16;
5715: else if ( obpe < 32 ) nd_bpe = 32;
5716: else error("nd_det_reconstruct : exponent too large");
5717:
5718: nd_setup_parameters(nd_nvar,0);
5719: prev_nm_free_list = _nm_free_list;
5720: _nm_free_list = 0;
5721: for ( k = j; k < n; k++ )
5722: for (l = j; l < n; l++ )
5723: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5724: ndv_realloc(d,obpe,oadv,oepos);
5725: prev_nm_free_list = 0;
5726: #if 0
5727: GC_gcollect();
5728: #endif
5729: }
5730:
5731: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5732: {
5733: UINT *d0,*d1,*d,*t,*r;
5734: int k,l;
5735:
5736: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5737: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5738: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5739: for ( k = j; k < n; k++ )
5740: for ( l = j; l < n; l++ )
5741: if ( dm[k][l] ) {
5742: d = ndv_compute_bound(dm[k][l]);
5743: ndl_lcm(d,d0,d1);
5744: t = d1; d1 = d0; d0 = t;
5745: }
5746: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5747: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5748: return r;
1.117 noro 5749: }
5750:
5751: DL nd_separate_d(UINT *d,UINT *trans)
5752: {
1.121 noro 5753: int n,td,i,e,j;
1.117 noro 5754: DL a;
5755:
5756: ndl_zero(trans);
5757: td = 0;
1.121 noro 5758: for ( i = 0; i < nd_ntrans; i++ ) {
5759: e = GET_EXP(d,i);
5760: PUT_EXP(trans,i,e);
5761: td += MUL_WEIGHT(e,i);
5762: }
5763: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5764: /* homogenized */
5765: i = nd_nvar-1;
1.117 noro 5766: e = GET_EXP(d,i);
5767: PUT_EXP(trans,i,e);
5768: td += MUL_WEIGHT(e,i);
5769: }
5770: TD(trans) = td;
5771: if ( nd_blockmask) ndl_weight_mask(trans);
5772: NEWDL(a,nd_nalg);
5773: td = 0;
1.121 noro 5774: for ( i = 0; i < nd_nalg; i++ ) {
5775: j = nd_ntrans+i;
5776: e = GET_EXP(d,j);
5777: a->d[i] = e;
1.117 noro 5778: td += e;
5779: }
5780: a->td = td;
5781: return a;
5782: }
5783:
1.118 noro 5784: int nd_monic(int mod,ND *p)
1.117 noro 5785: {
5786: UINT *trans,*t;
5787: DL alg;
5788: MP mp0,mp;
1.118 noro 5789: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5790: ND r;
1.117 noro 5791: DL dl;
5792: DP nm;
5793: NDV ndv;
1.118 noro 5794: DAlg inv,cd;
1.117 noro 5795: ND s,c;
1.118 noro 5796: Q l,mul;
5797: N ln;
5798: int n,ntrans,i,e,td,is_lc,len;
5799: NumberField nf;
5800: struct oEGT eg0,eg1;
5801:
5802: if ( !(nf = get_numberfield()) )
5803: error("nd_monic : current_numberfield is not set");
1.117 noro 5804:
1.118 noro 5805: /* Q coef -> DAlg coef */
5806: NEWNM(ma0); ma = ma0;
5807: m = BDY(*p);
5808: is_lc = 1;
5809: while ( 1 ) {
5810: NEWMP(mp0); mp = mp0;
5811: mp->c = (P)CQ(m);
5812: mp->dl = nd_separate_d(DL(m),DL(ma));
5813: NEWNM(mb);
5814: for ( m = NEXT(m); m; m = NEXT(m) ) {
5815: alg = nd_separate_d(DL(m),DL(mb));
5816: if ( !ndl_equal(DL(ma),DL(mb)) )
5817: break;
5818: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5819: }
5820: NEXT(mp) = 0;
5821: MKDP(nd_nalg,mp0,nm);
5822: MKDAlg(nm,ONE,cd);
5823: if ( is_lc == 1 ) {
5824: /* if the lc is a rational number, we have nothing to do */
5825: if ( !mp0->dl->td )
5826: return 1;
5827:
5828: get_eg(&eg0);
5829: invdalg(cd,&inv);
5830: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5831: /* check the validity of inv */
5832: if ( mod && !rem(NM(inv->dn),mod) )
5833: return 0;
5834: CA(ma) = nf->one;
5835: is_lc = 0;
5836: ln = ONEN;
5837: } else {
5838: muldalg(cd,inv,&CA(ma));
5839: lcmn(ln,NM(CA(ma)->dn),&ln);
5840: }
5841: if ( m ) {
5842: NEXT(ma) = mb; ma = mb;
5843: } else {
5844: NEXT(ma) = 0;
5845: break;
5846: }
5847: }
5848: /* l = lcm(denoms) */
5849: NTOQ(ln,1,l);
5850: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5851: divq(l,CA(m)->dn,&mul);
5852: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5853: NEXTNM(mr0,mr);
5854: mulq((Q)mp->c,mul,&CQ(mr));
5855: dl = mp->dl;
5856: td = TD(DL(m));
5857: ndl_copy(DL(m),DL(mr));
1.121 noro 5858: for ( i = 0; i < nd_nalg; i++ ) {
5859: e = dl->d[i];
5860: PUT_EXP(DL(mr),i+nd_ntrans,e);
5861: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5862: }
5863: TD(DL(mr)) = td;
5864: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5865: }
1.117 noro 5866: }
1.118 noro 5867: NEXT(mr) = 0;
5868: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5869: MKND(NV(*p),mr0,len,r);
5870: /* XXX */
5871: SG(r) = SG(*p);
5872: nd_free(*p);
5873: *p = r;
5874: return 1;
1.59 noro 5875: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>