![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to nd.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [local] / OpenXM_contrib2 / asir2000 / engine|
Return to nd.c CVS log | Up to [local] / OpenXM_contrib2 / asir2000 / engine |
1.148 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.147 2006/11/27 07:54:41 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:
1.146 noro 12: static NODE nd_subst;
13: static VL nd_vc;
1.121 noro 14: static int nd_ntrans;
1.117 noro 15: static int nd_nalg;
1.103 noro 16: #if 0
1.74 noro 17: static int ndv_alloc;
1.103 noro 18: #endif
1.87 noro 19: #if 1
1.69 noro 20: static int nd_f4_nsp=0x7fffffff;
1.87 noro 21: #else
22: static int nd_f4_nsp=50;
23: #endif
1.42 noro 24: static double nd_scale=2;
1.61 noro 25: static UINT **nd_bound;
1.42 noro 26: static struct order_spec *nd_ord;
27: static EPOS nd_epos;
1.43 noro 28: static BlockMask nd_blockmask;
1.42 noro 29: static int nd_nvar;
30: static int nd_isrlex;
31: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 32: static UINT nd_mask[32];
33: static UINT nd_mask0,nd_mask1;
1.42 noro 34:
1.20 noro 35: static NDV *nd_ps;
1.53 noro 36: static NDV *nd_ps_trace;
1.42 noro 37: static RHist *nd_psh;
38: static int nd_psn,nd_pslen;
39: static RHist *nd_red;
1.96 noro 40: static int *nd_work_vector;
41: static int **nd_matrix;
42: static int nd_matrix_len;
1.97 noro 43: static struct weight_or_block *nd_worb;
44: static int nd_worb_len;
1.42 noro 45: static int nd_found,nd_create,nd_notfirst;
46: static int nmv_adv;
1.77 noro 47: static int nd_demand;
1.1 noro 48:
1.119 noro 49: NumberField get_numberfield();
1.114 noro 50: UINT *nd_det_compute_bound(NDV **dm,int n,int j);
51: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d);
1.118 noro 52: int nd_monic(int m,ND *p);
1.129 noro 53: NDV plain_vect_to_ndv_q(Q *mat,int col,UINT *s0vect);
1.114 noro 54:
1.1 noro 55: void nd_free_private_storage()
56: {
57: _nm_free_list = 0;
1.5 noro 58: _ndp_free_list = 0;
1.71 noro 59: #if 0
1.1 noro 60: GC_gcollect();
1.71 noro 61: #endif
1.1 noro 62: }
63:
64: void _NM_alloc()
65: {
66: NM p;
67: int i;
68:
1.11 noro 69: for ( i = 0; i < 1024; i++ ) {
1.61 noro 70: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 71: p->next = _nm_free_list; _nm_free_list = p;
72: }
73: }
74:
75: void _ND_alloc()
76: {
77: ND p;
78: int i;
79:
80: for ( i = 0; i < 1024; i++ ) {
81: p = (ND)GC_malloc(sizeof(struct oND));
82: p->body = (NM)_nd_free_list; _nd_free_list = p;
83: }
84: }
85:
86: void _NDP_alloc()
87: {
88: ND_pairs p;
89: int i;
90:
1.11 noro 91: for ( i = 0; i < 1024; i++ ) {
1.1 noro 92: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 93: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 94: p->next = _ndp_free_list; _ndp_free_list = p;
95: }
96: }
97:
1.30 noro 98: INLINE int nd_length(ND p)
1.1 noro 99: {
100: NM m;
101: int i;
102:
103: if ( !p )
104: return 0;
105: else {
106: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
107: return i;
108: }
109: }
110:
1.61 noro 111: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 112: {
1.61 noro 113: UINT u1,u2;
1.1 noro 114: int i,j;
115:
1.34 noro 116: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 117: #if USE_UNROLL
1.1 noro 118: switch ( nd_bpe ) {
1.62 noro 119: case 3:
120: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
121: u1 = d1[i]; u2 = d2[i];
122: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
123: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
124: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
125: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
126: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
127: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
128: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
129: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
130: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
131: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
132: }
133: return 1;
134: break;
1.1 noro 135: case 4:
1.41 noro 136: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 137: u1 = d1[i]; u2 = d2[i];
138: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 139: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
140: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
141: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
142: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
143: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
144: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
145: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 146: }
147: return 1;
148: break;
149: case 6:
1.41 noro 150: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 151: u1 = d1[i]; u2 = d2[i];
152: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 153: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
154: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
155: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
156: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 157: }
158: return 1;
159: break;
160: case 8:
1.41 noro 161: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 162: u1 = d1[i]; u2 = d2[i];
163: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 164: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
165: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
166: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 167: }
168: return 1;
169: break;
170: case 16:
1.41 noro 171: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 172: u1 = d1[i]; u2 = d2[i];
173: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 174: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 175: }
176: return 1;
177: break;
178: case 32:
1.41 noro 179: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 180: if ( d1[i] < d2[i] ) return 0;
181: return 1;
182: break;
183: default:
1.41 noro 184: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 185: u1 = d1[i]; u2 = d2[i];
186: for ( j = 0; j < nd_epw; j++ )
187: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
188: }
189: return 1;
190: }
1.65 noro 191: #else
192: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
193: u1 = d1[i]; u2 = d2[i];
194: for ( j = 0; j < nd_epw; j++ )
195: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
196: }
197: return 1;
198: #endif
1.1 noro 199: }
200:
1.61 noro 201: /*
202: * If the current order is a block order,
203: * then the last block is length 1 and contains
204: * the homo variable. Otherwise, the original
205: * order is either 0 or 2.
206: */
207:
208: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 209: {
1.61 noro 210: int w,i,e,n,omask0;
211:
212: omask0 = (1<<obpe)-1;
213: n = nd_nvar-1;
214: ndl_zero(r);
215: for ( i = 0; i < n; i++ ) {
216: e = GET_EXP_OLD(d,i);
217: PUT_EXP(r,i,e);
218: }
219: w = TD(d);
220: PUT_EXP(r,nd_nvar-1,weight-w);
221: TD(r) = weight;
222: if ( nd_blockmask ) ndl_weight_mask(r);
223: }
224:
225: void ndl_dehomogenize(UINT *d)
226: {
227: UINT mask;
228: UINT h;
1.31 noro 229: int i,bits;
1.23 noro 230:
1.44 noro 231: if ( nd_blockmask ) {
232: h = GET_EXP(d,nd_nvar-1);
1.45 noro 233: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 234: TD(d) -= h;
235: d[nd_exporigin-1] -= h;
236: } else {
237: if ( nd_isrlex ) {
238: if ( nd_bpe == 32 ) {
239: h = d[nd_exporigin];
240: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
241: d[i-1] = d[i];
242: d[i-1] = 0;
243: TD(d) -= h;
244: } else {
245: bits = nd_epw*nd_bpe;
246: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
247: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
248: for ( i = nd_exporigin; i < nd_wpd; i++ )
249: d[i] = ((d[i]<<nd_bpe)&mask)
250: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
251: TD(d) -= h;
252: }
1.45 noro 253: } else {
254: h = GET_EXP(d,nd_nvar-1);
255: XOR_EXP(d,nd_nvar-1,h);
256: TD(d) -= h;
257: }
1.44 noro 258: }
1.23 noro 259: }
260:
1.61 noro 261: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 262: {
1.61 noro 263: UINT t1,t2,u,u1,u2;
1.43 noro 264: int i,j,l;
1.1 noro 265:
1.65 noro 266: #if USE_UNROLL
1.1 noro 267: switch ( nd_bpe ) {
1.62 noro 268: case 3:
269: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
270: u1 = d1[i]; u2 = d2[i];
271: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
272: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
273: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
274: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
275: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
276: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
277: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
278: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
279: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
280: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
281: d[i] = u;
282: }
283: break;
1.1 noro 284: case 4:
1.41 noro 285: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 286: u1 = d1[i]; u2 = d2[i];
287: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 288: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
289: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
290: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
291: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
292: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
293: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
294: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 295: d[i] = u;
296: }
297: break;
298: case 6:
1.41 noro 299: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 300: u1 = d1[i]; u2 = d2[i];
301: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 302: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
303: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
304: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
305: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 306: d[i] = u;
307: }
308: break;
309: case 8:
1.41 noro 310: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 311: u1 = d1[i]; u2 = d2[i];
312: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 313: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
314: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
315: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 316: d[i] = u;
317: }
318: break;
319: case 16:
1.41 noro 320: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 321: u1 = d1[i]; u2 = d2[i];
322: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 323: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 324: d[i] = u;
325: }
326: break;
327: case 32:
1.41 noro 328: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 329: u1 = d1[i]; u2 = d2[i];
330: d[i] = u1>u2?u1:u2;
331: }
332: break;
333: default:
1.41 noro 334: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 335: u1 = d1[i]; u2 = d2[i];
336: for ( j = 0, u = 0; j < nd_epw; j++ ) {
337: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
338: }
339: d[i] = u;
340: }
341: break;
342: }
1.65 noro 343: #else
344: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
345: u1 = d1[i]; u2 = d2[i];
346: for ( j = 0, u = 0; j < nd_epw; j++ ) {
347: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
348: }
349: d[i] = u;
350: }
351: #endif
1.39 noro 352: TD(d) = ndl_weight(d);
1.61 noro 353: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 354: }
355:
1.61 noro 356: int ndl_weight(UINT *d)
1.1 noro 357: {
1.61 noro 358: UINT t,u;
1.1 noro 359: int i,j;
360:
1.60 noro 361: if ( current_dl_weight_vector )
362: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
363: u = GET_EXP(d,i);
364: t += MUL_WEIGHT(u,i);
365: }
366: else
367: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
368: u = d[i];
369: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
370: t += (u&nd_mask0);
371: }
1.1 noro 372: return t;
373: }
374:
1.61 noro 375: void ndl_weight_mask(UINT *d)
1.43 noro 376: {
1.61 noro 377: UINT t,u;
378: UINT *mask;
379: int i,j,k,l;
1.43 noro 380:
1.61 noro 381: l = nd_blockmask->n;
382: for ( k = 0; k < l; k++ ) {
383: mask = nd_blockmask->mask[k];
384: if ( current_dl_weight_vector )
385: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
386: u = GET_EXP_MASK(d,i,mask);
387: t += MUL_WEIGHT(u,i);
388: }
389: else
390: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
391: u = d[i]&mask[i];
392: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
393: t += (u&nd_mask0);
394: }
395: d[k+1] = t;
396: }
1.43 noro 397: }
398:
1.61 noro 399: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 400: {
401: int i;
402:
1.41 noro 403: d1 += nd_exporigin;
404: d2 += nd_exporigin;
405: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 406: if ( *d1 > *d2 )
1.32 noro 407: return nd_isrlex ? -1 : 1;
1.1 noro 408: else if ( *d1 < *d2 )
1.32 noro 409: return nd_isrlex ? 1 : -1;
1.1 noro 410: return 0;
411: }
412:
1.61 noro 413: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 414: {
415: int i,l,j,ord_o,ord_l;
416: struct order_pair *op;
1.61 noro 417: UINT t1,t2,m;
418: UINT *mask;
1.43 noro 419:
420: l = nd_blockmask->n;
421: op = nd_blockmask->order_pair;
422: for ( j = 0; j < l; j++ ) {
423: mask = nd_blockmask->mask[j];
424: ord_o = op[j].order;
425: if ( ord_o < 2 )
1.44 noro 426: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
427: else if ( t1 < t2 ) return -1;
1.43 noro 428: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 429: m = mask[i];
430: t1 = d1[i]&m;
431: t2 = d2[i]&m;
1.43 noro 432: if ( t1 > t2 )
433: return !ord_o ? -1 : 1;
434: else if ( t1 < t2 )
435: return !ord_o ? 1 : -1;
436: }
437: }
438: return 0;
439: }
440:
1.96 noro 441: int ndl_matrix_compare(UINT *d1,UINT *d2)
442: {
443: int i,j,s;
444: int *v;
445:
446: for ( j = 0; j < nd_nvar; j++ )
447: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
448: for ( i = 0; i < nd_matrix_len; i++ ) {
449: v = nd_matrix[i];
450: for ( j = 0, s = 0; j < nd_nvar; j++ )
451: s += v[j]*nd_work_vector[j];
452: if ( s > 0 ) return 1;
453: else if ( s < 0 ) return -1;
454: }
455: return 0;
456: }
457:
1.97 noro 458: int ndl_composite_compare(UINT *d1,UINT *d2)
459: {
460: int i,j,s,start,end,len,o;
461: int *v;
462: struct sparse_weight *sw;
463:
464: for ( j = 0; j < nd_nvar; j++ )
465: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
466: for ( i = 0; i < nd_worb_len; i++ ) {
467: len = nd_worb[i].length;
468: switch ( nd_worb[i].type ) {
469: case IS_DENSE_WEIGHT:
470: v = nd_worb[i].body.dense_weight;
471: for ( j = 0, s = 0; j < len; j++ )
472: s += v[j]*nd_work_vector[j];
473: if ( s > 0 ) return 1;
474: else if ( s < 0 ) return -1;
475: break;
476: case IS_SPARSE_WEIGHT:
477: sw = nd_worb[i].body.sparse_weight;
478: for ( j = 0, s = 0; j < len; j++ )
479: s += sw[j].value*nd_work_vector[sw[j].pos];
480: if ( s > 0 ) return 1;
481: else if ( s < 0 ) return -1;
482: break;
483: case IS_BLOCK:
484: o = nd_worb[i].body.block.order;
485: start = nd_worb[i].body.block.start;
486: switch ( o ) {
487: case 0:
488: end = start+len;
489: for ( j = start, s = 0; j < end; j++ )
490: s += MUL_WEIGHT(nd_work_vector[j],j);
491: if ( s > 0 ) return 1;
492: else if ( s < 0 ) return -1;
493: for ( j = end-1; j >= start; j-- )
494: if ( nd_work_vector[j] < 0 ) return 1;
495: else if ( nd_work_vector[j] > 0 ) return -1;
496: break;
497: case 1:
498: end = start+len;
499: for ( j = start, s = 0; j < end; j++ )
500: s += MUL_WEIGHT(nd_work_vector[j],j);
501: if ( s > 0 ) return 1;
502: else if ( s < 0 ) return -1;
503: for ( j = start; j < end; j++ )
504: if ( nd_work_vector[j] > 0 ) return 1;
505: else if ( nd_work_vector[j] < 0 ) return -1;
506: break;
507: case 2:
508: for ( j = start; j < end; j++ )
509: if ( nd_work_vector[j] > 0 ) return 1;
510: else if ( nd_work_vector[j] < 0 ) return -1;
511: break;
512: }
513: break;
514: }
515: }
516: return 0;
517: }
518:
1.58 noro 519: /* TDH -> WW -> TD-> RL */
520:
1.61 noro 521: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 522: {
523: int i,m,e1,e2;
524:
525: if ( TD(d1) > TD(d2) ) return 1;
526: else if ( TD(d1) < TD(d2) ) return -1;
527: m = nd_nvar>>1;
528: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
529: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
530: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
531: }
532: if ( e1 > e2 ) return 1;
533: else if ( e1 < e2 ) return -1;
534: return ndl_lex_compare(d1,d2);
535: }
536:
1.61 noro 537: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 538: {
539: int i;
540:
1.81 noro 541: switch ( nd_wpd ) {
542: case 2:
543: if ( TD(d2) != TD(d1) ) return 0;
544: if ( d2[1] != d1[1] ) return 0;
545: return 1;
546: break;
547: case 3:
548: if ( TD(d2) != TD(d1) ) return 0;
549: if ( d2[1] != d1[1] ) return 0;
550: if ( d2[2] != d1[2] ) return 0;
551: return 1;
552: break;
553: default:
554: for ( i = 0; i < nd_wpd; i++ )
555: if ( *d1++ != *d2++ ) return 0;
556: return 1;
557: break;
558: }
1.1 noro 559: }
560:
1.61 noro 561: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 562: {
563: int i;
564:
565: switch ( nd_wpd ) {
1.41 noro 566: case 2:
1.34 noro 567: TD(d2) = TD(d1);
568: d2[1] = d1[1];
1.6 noro 569: break;
1.41 noro 570: case 3:
1.34 noro 571: TD(d2) = TD(d1);
1.6 noro 572: d2[1] = d1[1];
1.34 noro 573: d2[2] = d1[2];
1.6 noro 574: break;
575: default:
1.41 noro 576: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 577: d2[i] = d1[i];
578: break;
579: }
580: }
581:
1.61 noro 582: INLINE void ndl_zero(UINT *d)
583: {
584: int i;
585: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
586: }
587:
588: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 589: {
590: int i;
591:
1.43 noro 592: #if 1
1.6 noro 593: switch ( nd_wpd ) {
1.41 noro 594: case 2:
595: TD(d) = TD(d1)+TD(d2);
1.34 noro 596: d[1] = d1[1]+d2[1];
1.6 noro 597: break;
1.41 noro 598: case 3:
599: TD(d) = TD(d1)+TD(d2);
1.6 noro 600: d[1] = d1[1]+d2[1];
1.34 noro 601: d[2] = d1[2]+d2[2];
1.6 noro 602: break;
603: default:
1.43 noro 604: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 605: break;
606: }
1.43 noro 607: #else
608: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
609: #endif
1.6 noro 610: }
611:
1.55 noro 612: /* d1 += d2 */
1.61 noro 613: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 614: {
615: int i;
616:
617: #if 1
618: switch ( nd_wpd ) {
619: case 2:
620: TD(d1) += TD(d2);
621: d1[1] += d2[1];
622: break;
623: case 3:
624: TD(d1) += TD(d2);
625: d1[1] += d2[1];
626: d1[2] += d2[2];
627: break;
628: default:
629: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
630: break;
631: }
632: #else
633: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
634: #endif
635: }
636:
1.61 noro 637: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 638: {
639: int i;
640:
1.43 noro 641: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 642: }
643:
1.61 noro 644: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 645: {
1.61 noro 646: UINT t1,t2,u,u1,u2;
1.1 noro 647: int i,j;
648:
1.65 noro 649: #if USE_UNROLL
1.1 noro 650: switch ( nd_bpe ) {
1.62 noro 651: case 3:
652: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
653: u1 = d1[i]; u2 = d2[i];
654: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
655: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
656: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
657: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
658: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
659: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
660: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
661: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
662: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
663: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
664: }
665: return 1;
666: break;
1.1 noro 667: case 4:
1.41 noro 668: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 669: u1 = d1[i]; u2 = d2[i];
670: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 671: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
672: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
673: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
674: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
675: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
676: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
677: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 678: }
679: return 1;
680: break;
681: case 6:
1.41 noro 682: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 683: u1 = d1[i]; u2 = d2[i];
684: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 685: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
686: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
687: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
688: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 689: }
690: return 1;
691: break;
692: case 8:
1.41 noro 693: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 694: u1 = d1[i]; u2 = d2[i];
695: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 696: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
697: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
698: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 699: }
700: return 1;
701: break;
702: case 16:
1.41 noro 703: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 704: u1 = d1[i]; u2 = d2[i];
705: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 706: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 707: }
708: return 1;
709: break;
710: case 32:
1.41 noro 711: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 712: if ( d1[i] && d2[i] ) return 0;
713: return 1;
714: break;
715: default:
1.41 noro 716: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 717: u1 = d1[i]; u2 = d2[i];
718: for ( j = 0; j < nd_epw; j++ ) {
719: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
720: u1 >>= nd_bpe; u2 >>= nd_bpe;
721: }
722: }
723: return 1;
724: break;
725: }
1.65 noro 726: #else
727: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
728: u1 = d1[i]; u2 = d2[i];
729: for ( j = 0; j < nd_epw; j++ ) {
730: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
731: u1 >>= nd_bpe; u2 >>= nd_bpe;
732: }
733: }
734: return 1;
735: #endif
1.1 noro 736: }
737:
1.114 noro 738: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 739: {
1.61 noro 740: UINT u2;
1.5 noro 741: int i,j,ind,k;
1.1 noro 742:
1.5 noro 743: ind = 0;
1.65 noro 744: #if USE_UNROLL
1.5 noro 745: switch ( nd_bpe ) {
1.62 noro 746: case 3:
747: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
748: u2 = d2[i];
749: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
750: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
751: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
752: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
753: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
754: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
755: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
756: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
757: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
758: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
759: }
760: return 0;
761: break;
1.5 noro 762: case 4:
1.41 noro 763: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 764: u2 = d2[i];
765: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
766: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
767: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
768: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
769: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
770: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
771: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
772: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
773: }
774: return 0;
775: break;
776: case 6:
1.41 noro 777: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 778: u2 = d2[i];
779: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
780: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
781: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
782: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
783: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
784: }
785: return 0;
786: break;
787: case 8:
1.41 noro 788: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 789: u2 = d2[i];
790: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
791: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
792: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
793: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
794: }
795: return 0;
796: break;
797: case 16:
1.41 noro 798: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 799: u2 = d2[i];
800: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
801: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
802: }
803: return 0;
804: break;
805: case 32:
1.41 noro 806: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 807: if ( d1[i]+d2[i]<d1[i] ) return 1;
808: return 0;
809: break;
810: default:
1.41 noro 811: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 812: u2 = d2[i];
813: k = (nd_epw-1)*nd_bpe;
814: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
815: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
816: }
817: return 0;
818: break;
819: }
1.65 noro 820: #else
821: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
822: u2 = d2[i];
823: k = (nd_epw-1)*nd_bpe;
824: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
825: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
826: }
827: return 0;
828: #endif
1.1 noro 829: }
830:
1.114 noro 831: int ndl_check_bound2(int index,UINT *d2)
832: {
833: return ndl_check_bound(nd_bound[index],d2);
834: }
835:
1.61 noro 836: INLINE int ndl_hash_value(UINT *d)
1.1 noro 837: {
838: int i;
839: int r;
840:
1.34 noro 841: r = 0;
1.41 noro 842: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 843: r = ((r<<16)+d[i])%REDTAB_LEN;
844: return r;
845: }
846:
1.63 noro 847: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 848: {
1.13 noro 849: RHist r;
1.6 noro 850: int d,k,i;
1.1 noro 851:
1.63 noro 852: d = ndl_hash_value(dg);
1.13 noro 853: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 854: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 855: if ( k > 0 ) nd_notfirst++;
856: nd_found++;
1.13 noro 857: return r->index;
1.1 noro 858: }
859: }
1.13 noro 860: if ( Reverse )
861: for ( i = nd_psn-1; i >= 0; i-- ) {
862: r = nd_psh[i];
1.34 noro 863: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 864: nd_create++;
1.34 noro 865: nd_append_red(dg,i);
1.13 noro 866: return i;
867: }
868: }
869: else
870: for ( i = 0; i < nd_psn; i++ ) {
871: r = nd_psh[i];
1.34 noro 872: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 873: nd_create++;
1.34 noro 874: nd_append_red(dg,i);
1.13 noro 875: return i;
876: }
1.1 noro 877: }
1.6 noro 878: return -1;
1.1 noro 879: }
880:
1.63 noro 881: ND nd_merge(ND p1,ND p2)
882: {
883: int n,c;
884: int t,can,td1,td2;
885: ND r;
886: NM m1,m2,mr0,mr,s;
887:
888: if ( !p1 ) return p2;
889: else if ( !p2 ) return p1;
890: else {
891: can = 0;
892: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
893: c = DL_COMPARE(DL(m1),DL(m2));
894: switch ( c ) {
895: case 0:
896: s = m1; m1 = NEXT(m1);
897: can++; NEXTNM2(mr0,mr,s);
898: s = m2; m2 = NEXT(m2); FREENM(s);
899: break;
900: case 1:
901: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
902: break;
903: case -1:
904: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
905: break;
906: }
907: }
908: if ( !mr0 )
909: if ( m1 ) mr0 = m1;
910: else if ( m2 ) mr0 = m2;
911: else return 0;
912: else if ( m1 ) NEXT(mr) = m1;
913: else if ( m2 ) NEXT(mr) = m2;
914: else NEXT(mr) = 0;
915: BDY(p1) = mr0;
916: SG(p1) = MAX(SG(p1),SG(p2));
917: LEN(p1) = LEN(p1)+LEN(p2)-can;
918: FREEND(p2);
919: return p1;
920: }
921: }
922:
1.31 noro 923: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 924: {
925: int n,c;
1.34 noro 926: int t,can,td1,td2;
1.1 noro 927: ND r;
928: NM m1,m2,mr0,mr,s;
929:
1.34 noro 930: if ( !p1 ) return p2;
931: else if ( !p2 ) return p1;
1.71 noro 932: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.113 noro 933: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 934: else {
1.30 noro 935: can = 0;
1.1 noro 936: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 937: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 938: switch ( c ) {
939: case 0:
1.19 noro 940: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 941: if ( t < 0 ) t += mod;
1.1 noro 942: s = m1; m1 = NEXT(m1);
943: if ( t ) {
1.34 noro 944: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 945: } else {
1.34 noro 946: can += 2; FREENM(s);
1.17 noro 947: }
948: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 949: break;
950: case 1:
951: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
952: break;
953: case -1:
954: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
955: break;
956: }
957: }
958: if ( !mr0 )
959: if ( m1 ) mr0 = m1;
960: else if ( m2 ) mr0 = m2;
961: else return 0;
962: else if ( m1 ) NEXT(mr) = m1;
963: else if ( m2 ) NEXT(mr) = m2;
964: else NEXT(mr) = 0;
965: BDY(p1) = mr0;
966: SG(p1) = MAX(SG(p1),SG(p2));
967: LEN(p1) = LEN(p1)+LEN(p2)-can;
968: FREEND(p2);
969: return p1;
970: }
971: }
972:
973: /* XXX on opteron, the inlined manipulation of destructive additon of
974: * two NM seems to make gcc optimizer get confused, so the part is
975: * done in a function.
976: */
977:
1.113 noro 978: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 979: {
980: NM s;
1.146 noro 981: P t;
1.95 noro 982: int can;
983:
1.146 noro 984: addp(nd_vc,CP(*m1),CP(*m2),&t);
1.95 noro 985: s = *m1; *m1 = NEXT(*m1);
986: if ( t ) {
1.146 noro 987: can = 1; NEXTNM2(*mr0,*mr,s); CP(*mr) = (t);
1.95 noro 988: } else {
989: can = 2; FREENM(s);
990: }
991: s = *m2; *m2 = NEXT(*m2); FREENM(s);
992: return can;
993: }
994:
1.113 noro 995: ND nd_add_q(ND p1,ND p2)
1.95 noro 996: {
997: int n,c,can;
998: ND r;
999: NM m1,m2,mr0,mr,s;
1.146 noro 1000: P t;
1.95 noro 1001:
1002: if ( !p1 ) return p2;
1003: else if ( !p2 ) return p1;
1004: else {
1005: can = 0;
1006: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1007: c = DL_COMPARE(DL(m1),DL(m2));
1008: switch ( c ) {
1009: case 0:
1010: #if defined(__x86_64__)
1.113 noro 1011: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1012: #else
1.146 noro 1013: addp(nd_vc,CP(m1),CP(m2),&t);
1.95 noro 1014: s = m1; m1 = NEXT(m1);
1015: if ( t ) {
1.146 noro 1016: can++; NEXTNM2(mr0,mr,s); CP(mr) = (t);
1.95 noro 1017: } else {
1018: can += 2; FREENM(s);
1019: }
1020: s = m2; m2 = NEXT(m2); FREENM(s);
1021: #endif
1.17 noro 1022: break;
1023: case 1:
1024: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1025: break;
1026: case -1:
1027: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1028: break;
1029: }
1030: }
1031: if ( !mr0 )
1.34 noro 1032: if ( m1 ) mr0 = m1;
1033: else if ( m2 ) mr0 = m2;
1034: else return 0;
1035: else if ( m1 ) NEXT(mr) = m1;
1036: else if ( m2 ) NEXT(mr) = m2;
1037: else NEXT(mr) = 0;
1.17 noro 1038: BDY(p1) = mr0;
1039: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1040: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1041: FREEND(p2);
1042: return p1;
1043: }
1044: }
1045:
1.71 noro 1046: ND nd_add_sf(ND p1,ND p2)
1047: {
1048: int n,c,can;
1049: ND r;
1050: NM m1,m2,mr0,mr,s;
1051: int t;
1052:
1053: if ( !p1 ) return p2;
1054: else if ( !p2 ) return p1;
1055: else {
1056: can = 0;
1057: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1058: c = DL_COMPARE(DL(m1),DL(m2));
1059: switch ( c ) {
1060: case 0:
1061: t = _addsf(CM(m1),CM(m2));
1062: s = m1; m1 = NEXT(m1);
1063: if ( t ) {
1064: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1065: } else {
1066: can += 2; FREENM(s);
1067: }
1068: s = m2; m2 = NEXT(m2); FREENM(s);
1069: break;
1070: case 1:
1071: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1072: break;
1073: case -1:
1074: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1075: break;
1076: }
1077: }
1078: if ( !mr0 )
1079: if ( m1 ) mr0 = m1;
1080: else if ( m2 ) mr0 = m2;
1081: else return 0;
1082: else if ( m1 ) NEXT(mr) = m1;
1083: else if ( m2 ) NEXT(mr) = m2;
1084: else NEXT(mr) = 0;
1085: BDY(p1) = mr0;
1086: SG(p1) = MAX(SG(p1),SG(p2));
1087: LEN(p1) = LEN(p1)+LEN(p2)-can;
1088: FREEND(p2);
1089: return p1;
1090: }
1091: }
1092:
1.146 noro 1093: ND nd_reduce2(int mod,ND d,ND g,NDV p,NM mul,NDC dn)
1094: {
1095: int c,c1,c2;
1096: Q cg,cred,gcd,tq;
1097: P cgp,credp,gcdp;
1098: Obj tr,tr1;
1099:
1100: if ( mod == -1 )
1101: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1102: else if ( mod ) {
1103: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1104: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1105: } else if ( nd_vc ) {
1106: ezgcdpz(nd_vc,HCP(g),HCP(p),&gcdp);
1107: divsp(nd_vc,HCP(g),gcdp,&cgp); divsp(nd_vc,HCP(p),gcdp,&credp);
1108: chsgnp(cgp,&CP(mul));
1109: nd_mul_c_q(d,credp); nd_mul_c_q(g,credp);
1110: if ( dn ) {
1111: mulr(nd_vc,(Obj)dn->r,(Obj)credp,&tr);
1112: reductr(nd_vc,tr,&tr1); dn->r = (R)tr1;
1113: }
1114: } else {
1115: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1116: chsgnq(cg,&CQ(mul));
1117: nd_mul_c_q(d,(P)cred); nd_mul_c_q(g,(P)cred);
1118: if ( dn ) {
1119: mulq(dn->z,cred,&tq); dn->z = tq;
1120: }
1121: }
1122: return nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1123: }
1124:
1.1 noro 1125: /* ret=1 : success, ret=0 : overflow */
1.146 noro 1126: int nd_nf(int mod,ND d,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1127: {
1128: NM m,mrd,tail;
1.7 noro 1129: NM mul;
1.10 noro 1130: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1131: int c,c1,c2,dummy;
1.17 noro 1132: RHist h;
1.11 noro 1133: NDV p,red;
1.113 noro 1134: Q cg,cred,gcd,tq,qq;
1.21 noro 1135: double hmag;
1.146 noro 1136: P tp,tp1;
1137: Obj tr,tr1;
1.1 noro 1138:
1.113 noro 1139: if ( dn ) {
1140: if ( mod )
1141: dn->m = 1;
1.146 noro 1142: else if ( nd_vc )
1143: dn->r = (R)ONE;
1.113 noro 1144: else
1145: dn->z = ONE;
1146: }
1.1 noro 1147: if ( !g ) {
1.146 noro 1148: *rp = d;
1.1 noro 1149: return 1;
1150: }
1.146 noro 1151: if ( !mod ) hmag = ((double)p_mag(HCP(g)))*nd_scale;
1.21 noro 1152:
1.14 noro 1153: sugar0 = sugar = SG(g);
1.1 noro 1154: n = NV(g);
1.61 noro 1155: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.146 noro 1156: if ( d )
1157: for ( tail = BDY(d); NEXT(tail); tail = NEXT(tail) );
1158: for ( ; g; ) {
1.63 noro 1159: index = ndl_find_reducer(HDL(g));
1.6 noro 1160: if ( index >= 0 ) {
1.17 noro 1161: h = nd_psh[index];
1162: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1163: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1164: nd_free(g); nd_free(d);
1165: return 0;
1166: }
1.146 noro 1167: p = nd_demand ? ndv_load(index) : ps[index];
1168: g = nd_reduce2(mod,d,g,p,mul,dn);
1.34 noro 1169: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.146 noro 1170: if ( !mod && g && ((double)(p_mag(HCP(g))) > hmag) ) {
1171: tp = HCP(g);
1.21 noro 1172: nd_removecont2(d,g);
1.69 noro 1173: if ( dn ) {
1.146 noro 1174: if ( nd_vc ) {
1175: divsp(nd_vc,tp,HCP(g),&tp1);
1176: divr(nd_vc,(Obj)dn->r,(Obj)tp1,&tr);
1177: reductr(nd_vc,(Obj)tr,&tr1); dn->r = (R)tr1;
1178: } else {
1179: divq((Q)tp,HCQ(g),&qq);
1180: divq(dn->z,qq,&tq); dn->z = tq;
1181: }
1.69 noro 1182: }
1.146 noro 1183: hmag = ((double)p_mag(HCP(g)))*nd_scale;
1.21 noro 1184: }
1.1 noro 1185: } else if ( !full ) {
1186: *rp = g;
1187: return 1;
1188: } else {
1189: m = BDY(g);
1190: if ( NEXT(m) ) {
1.34 noro 1191: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1192: } else {
1193: FREEND(g); g = 0;
1194: }
1195: if ( d ) {
1.34 noro 1196: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1197: } else {
1.34 noro 1198: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1199: }
1200: }
1201: }
1.34 noro 1202: if ( d ) SG(d) = sugar;
1.1 noro 1203: *rp = d;
1204: return 1;
1205: }
1.28 noro 1206:
1.53 noro 1207: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1208: {
1209: int hindex,index;
1210: NDV p;
1211: ND u,d,red;
1212: NODE l;
1.31 noro 1213: NM mul,m,mrd,tail;
1.25 noro 1214: int sugar,psugar,n,h_reducible;
1215: PGeoBucket bucket;
1216: int c,c1,c2;
1.113 noro 1217: Q cg,cred,gcd,zzz;
1.25 noro 1218: RHist h;
1.28 noro 1219: double hmag,gmag;
1.77 noro 1220: int count = 0;
1221: int hcount = 0;
1.25 noro 1222:
1223: if ( !g ) {
1224: *rp = 0;
1225: return 1;
1226: }
1227: sugar = SG(g);
1228: n = NV(g);
1.113 noro 1229: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1230: bucket = create_pbucket();
1.31 noro 1231: add_pbucket(mod,bucket,g);
1.25 noro 1232: d = 0;
1.61 noro 1233: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1234: while ( 1 ) {
1.113 noro 1235: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1236: if ( hindex < 0 ) {
1.77 noro 1237: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1238: if ( d ) SG(d) = sugar;
1.25 noro 1239: *rp = d;
1240: return 1;
1241: }
1242: g = bucket->body[hindex];
1.63 noro 1243: index = ndl_find_reducer(HDL(g));
1.25 noro 1244: if ( index >= 0 ) {
1.77 noro 1245: count++;
1246: if ( !d ) hcount++;
1.25 noro 1247: h = nd_psh[index];
1248: ndl_sub(HDL(g),DL(h),DL(mul));
1249: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1250: nd_free(d);
1.25 noro 1251: free_pbucket(bucket);
1252: *rp = 0;
1253: return 0;
1254: }
1.53 noro 1255: p = ps[index];
1.71 noro 1256: if ( mod == -1 )
1257: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1258: else if ( mod ) {
1.25 noro 1259: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1260: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1261: } else {
1.113 noro 1262: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1263: chsgnq(cg,&CQ(mul));
1.146 noro 1264: nd_mul_c_q(d,(P)cred);
1.113 noro 1265: mulq_pbucket(bucket,cred);
1.26 noro 1266: g = bucket->body[hindex];
1.113 noro 1267: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1268: }
1.55 noro 1269: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1270: bucket->body[hindex] = nd_remove_head(g);
1271: red = nd_remove_head(red);
1.31 noro 1272: add_pbucket(mod,bucket,red);
1.34 noro 1273: psugar = SG(p)+TD(DL(mul));
1274: sugar = MAX(sugar,psugar);
1.28 noro 1275: if ( !mod && hmag && (gmag > hmag) ) {
1276: g = normalize_pbucket(mod,bucket);
1277: if ( !g ) {
1.34 noro 1278: if ( d ) SG(d) = sugar;
1.28 noro 1279: *rp = d;
1280: return 1;
1281: }
1282: nd_removecont2(d,g);
1.113 noro 1283: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1284: add_pbucket(mod,bucket,g);
1.28 noro 1285: }
1.25 noro 1286: } else if ( !full ) {
1287: g = normalize_pbucket(mod,bucket);
1.34 noro 1288: if ( g ) SG(g) = sugar;
1.25 noro 1289: *rp = g;
1290: return 1;
1291: } else {
1292: m = BDY(g);
1293: if ( NEXT(m) ) {
1.34 noro 1294: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1295: } else {
1296: FREEND(g); g = 0;
1297: }
1298: bucket->body[hindex] = g;
1299: NEXT(m) = 0;
1300: if ( d ) {
1.34 noro 1301: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1302: } else {
1.34 noro 1303: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1304: }
1305: }
1306: }
1307: }
1.27 noro 1308:
1.61 noro 1309: /* input : list of NDV, cand : list of NDV */
1.28 noro 1310:
1.61 noro 1311: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1312: {
1313: int n,i,stat;
1314: ND nf,d;
1.61 noro 1315: NDV r;
1.45 noro 1316: NODE t,s;
1.113 noro 1317: union oNDC dn;
1.45 noro 1318:
1.128 noro 1319: ndv_setup(0,0,cand,0);
1.31 noro 1320: n = length(cand);
1.28 noro 1321:
1322: /* membercheck : list is a subset of Id(cand) ? */
1323: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1324: again:
1.61 noro 1325: if ( nd_bpe > obpe )
1326: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1327: else
1328: r = (NDV)BDY(t);
1329: d = ndvtond(0,r);
1.146 noro 1330: stat = nd_nf(0,0,d,nd_ps,0,0,&nf);
1.45 noro 1331: if ( !stat ) {
1.103 noro 1332: nd_reconstruct(0,0);
1.45 noro 1333: goto again;
1334: } else if ( nf ) return 0;
1.71 noro 1335: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1336: }
1.71 noro 1337: if ( DP_Print ) { printf("\n"); }
1.28 noro 1338: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1339: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1340: /* XXX */
1.23 noro 1341: return 1;
1342: }
1.1 noro 1343:
1344: ND nd_remove_head(ND p)
1345: {
1346: NM m;
1347:
1348: m = BDY(p);
1349: if ( !NEXT(m) ) {
1.34 noro 1350: FREEND(p); p = 0;
1.31 noro 1351: } else {
1.34 noro 1352: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1353: }
1.1 noro 1354: FREENM(m);
1355: return p;
1356: }
1357:
1.69 noro 1358: ND nd_separate_head(ND p,ND *head)
1359: {
1360: NM m,m0;
1361: ND r;
1362:
1363: m = BDY(p);
1364: if ( !NEXT(m) ) {
1365: *head = p; p = 0;
1366: } else {
1367: m0 = m;
1368: BDY(p) = NEXT(m); LEN(p)--;
1369: NEXT(m0) = 0;
1370: MKND(NV(p),m0,1,r);
1371: *head = r;
1372: }
1373: return p;
1374: }
1375:
1.1 noro 1376: PGeoBucket create_pbucket()
1377: {
1378: PGeoBucket g;
1379:
1380: g = CALLOC(1,sizeof(struct oPGeoBucket));
1381: g->m = -1;
1382: return g;
1383: }
1384:
1.25 noro 1385: void free_pbucket(PGeoBucket b) {
1386: int i;
1387:
1.26 noro 1388: for ( i = 0; i <= b->m; i++ )
1.25 noro 1389: if ( b->body[i] ) {
1390: nd_free(b->body[i]);
1391: b->body[i] = 0;
1392: }
1393: GC_free(b);
1394: }
1395:
1.63 noro 1396: void add_pbucket_symbolic(PGeoBucket g,ND d)
1397: {
1398: int l,i,k,m;
1399:
1400: if ( !d )
1401: return;
1402: l = LEN(d);
1403: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1404: /* 2^(k-1) < l <= 2^k (=m) */
1405: d = nd_merge(g->body[k],d);
1406: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1407: g->body[k] = 0;
1408: d = nd_merge(g->body[k+1],d);
1409: }
1410: g->body[k] = d;
1411: g->m = MAX(g->m,k);
1412: }
1413:
1.31 noro 1414: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1415: {
1.31 noro 1416: int l,i,k,m;
1.1 noro 1417:
1.31 noro 1418: if ( !d )
1419: return;
1420: l = LEN(d);
1.29 noro 1421: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1422: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1423: d = nd_add(mod,g->body[k],d);
1424: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1425: g->body[k] = 0;
1.31 noro 1426: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1427: }
1428: g->body[k] = d;
1429: g->m = MAX(g->m,k);
1430: }
1431:
1.113 noro 1432: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1433: {
1434: int k;
1435:
1436: for ( k = 0; k <= g->m; k++ )
1.146 noro 1437: nd_mul_c_q(g->body[k],(P)c);
1.26 noro 1438: }
1439:
1.63 noro 1440: NM remove_head_pbucket_symbolic(PGeoBucket g)
1441: {
1442: int j,i,k,c;
1443: NM head;
1444:
1445: k = g->m;
1446: j = -1;
1447: for ( i = 0; i <= k; i++ ) {
1448: if ( !g->body[i] ) continue;
1449: if ( j < 0 ) j = i;
1450: else {
1451: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1452: if ( c > 0 )
1453: j = i;
1454: else if ( c == 0 )
1455: g->body[i] = nd_remove_head(g->body[i]);
1456: }
1457: }
1458: if ( j < 0 ) return 0;
1459: else {
1460: head = BDY(g->body[j]);
1461: if ( !NEXT(head) ) {
1462: FREEND(g->body[j]);
1463: g->body[j] = 0;
1464: } else {
1465: BDY(g->body[j]) = NEXT(head);
1466: LEN(g->body[j])--;
1467: }
1468: return head;
1469: }
1470: }
1471:
1.19 noro 1472: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1473: {
1474: int j,i,c,k,nv,sum;
1.61 noro 1475: UINT *di,*dj;
1.1 noro 1476: ND gi,gj;
1477:
1478: k = g->m;
1479: while ( 1 ) {
1480: j = -1;
1481: for ( i = 0; i <= k; i++ ) {
1482: if ( !(gi = g->body[i]) )
1483: continue;
1484: if ( j < 0 ) {
1485: j = i;
1486: gj = g->body[j];
1487: dj = HDL(gj);
1.14 noro 1488: sum = HCM(gj);
1.1 noro 1489: } else {
1.34 noro 1490: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1491: if ( c > 0 ) {
1.34 noro 1492: if ( sum ) HCM(gj) = sum;
1493: else g->body[j] = nd_remove_head(gj);
1.1 noro 1494: j = i;
1495: gj = g->body[j];
1496: dj = HDL(gj);
1.14 noro 1497: sum = HCM(gj);
1.1 noro 1498: } else if ( c == 0 ) {
1.71 noro 1499: if ( mod == -1 )
1500: sum = _addsf(sum,HCM(gi));
1501: else {
1502: sum = sum+HCM(gi)-mod;
1503: if ( sum < 0 ) sum += mod;
1504: }
1.1 noro 1505: g->body[i] = nd_remove_head(gi);
1506: }
1507: }
1508: }
1.34 noro 1509: if ( j < 0 ) return -1;
1.1 noro 1510: else if ( sum ) {
1.14 noro 1511: HCM(gj) = sum;
1.26 noro 1512: return j;
1.31 noro 1513: } else
1.26 noro 1514: g->body[j] = nd_remove_head(gj);
1515: }
1516: }
1517:
1.113 noro 1518: int head_pbucket_q(PGeoBucket g)
1.26 noro 1519: {
1520: int j,i,c,k,nv;
1.113 noro 1521: Q sum,t;
1.26 noro 1522: ND gi,gj;
1523:
1524: k = g->m;
1525: while ( 1 ) {
1526: j = -1;
1527: for ( i = 0; i <= k; i++ ) {
1.34 noro 1528: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1529: if ( j < 0 ) {
1530: j = i;
1531: gj = g->body[j];
1.113 noro 1532: sum = HCQ(gj);
1.26 noro 1533: } else {
1534: nv = NV(gi);
1.34 noro 1535: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1536: if ( c > 0 ) {
1.113 noro 1537: if ( sum ) HCQ(gj) = sum;
1.34 noro 1538: else g->body[j] = nd_remove_head(gj);
1.26 noro 1539: j = i;
1540: gj = g->body[j];
1.113 noro 1541: sum = HCQ(gj);
1.26 noro 1542: } else if ( c == 0 ) {
1.113 noro 1543: addq(sum,HCQ(gi),&t);
1544: sum = t;
1.26 noro 1545: g->body[i] = nd_remove_head(gi);
1546: }
1547: }
1548: }
1.34 noro 1549: if ( j < 0 ) return -1;
1.26 noro 1550: else if ( sum ) {
1.113 noro 1551: HCQ(gj) = sum;
1.1 noro 1552: return j;
1.31 noro 1553: } else
1.1 noro 1554: g->body[j] = nd_remove_head(gj);
1555: }
1556: }
1557:
1.25 noro 1558: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1559: {
1.31 noro 1560: int i;
1.1 noro 1561: ND r,t;
1562:
1563: r = 0;
1.28 noro 1564: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1565: r = nd_add(mod,r,g->body[i]);
1.28 noro 1566: g->body[i] = 0;
1567: }
1568: g->m = -1;
1.1 noro 1569: return r;
1570: }
1571:
1.122 noro 1572: int do_diagonalize(int sugar,int m)
1.92 noro 1573: {
1574: int i,nh,stat;
1575: NODE r,g,t;
1576: ND h,nf,s,head;
1577: NDV nfv;
1.113 noro 1578: Q q,num,den;
1.146 noro 1579: P nm,nmp,dn,mnp,dnp;
1.92 noro 1580:
1.143 noro 1581: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.92 noro 1582: if ( nd_demand )
1583: nfv = ndv_load(i);
1584: else
1585: nfv = nd_ps[i];
1586: s = ndvtond(m,nfv);
1587: s = nd_separate_head(s,&head);
1.146 noro 1588: stat = nd_nf(m,head,s,nd_ps,1,0,&nf);
1.122 noro 1589: if ( !stat ) return 0;
1.92 noro 1590: ndv_free(nfv);
1591: nd_removecont(m,nf);
1592: nfv = ndtondv(m,nf);
1593: nd_free(nf);
1594: nd_bound[i] = ndv_compute_bound(nfv);
1595: if ( nd_demand ) {
1596: ndv_save(nfv,i);
1597: ndv_free(nfv);
1598: } else
1599: nd_ps[i] = nfv;
1600: }
1.122 noro 1601: return 1;
1.92 noro 1602: }
1603:
1.27 noro 1604: /* return value = 0 => input is not a GB */
1605:
1.77 noro 1606: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1607: {
1608: int i,nh,sugar,stat;
1.23 noro 1609: NODE r,g,t;
1.1 noro 1610: ND_pairs d;
1611: ND_pairs l;
1.117 noro 1612: ND h,nf,s,head,nf1;
1.63 noro 1613: NDV nfv;
1.113 noro 1614: Q q,num,den;
1615: union oNDC dn;
1.140 noro 1616: int diag_count = 0;
1.1 noro 1617:
1.23 noro 1618: g = 0; d = 0;
1619: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1620: d = update_pairs(d,g,i);
1621: g = update_base(g,i);
1622: }
1623: sugar = 0;
1624: while ( d ) {
1625: again:
1626: l = nd_minp(d,&d);
1.14 noro 1627: if ( SG(l) != sugar ) {
1.122 noro 1628: if ( ishomo ) {
1.140 noro 1629: diag_count = 0;
1.122 noro 1630: stat = do_diagonalize(sugar,m);
1631: if ( !stat ) {
1632: NEXT(l) = d; d = l;
1633: d = nd_reconstruct(0,d);
1634: goto again;
1635: }
1636: }
1.14 noro 1637: sugar = SG(l);
1.71 noro 1638: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1639: }
1.53 noro 1640: stat = nd_sp(m,0,l,&h);
1.1 noro 1641: if ( !stat ) {
1642: NEXT(l) = d; d = l;
1.103 noro 1643: d = nd_reconstruct(0,d);
1.1 noro 1644: goto again;
1645: }
1.41 noro 1646: #if USE_GEOBUCKET
1.146 noro 1647: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1648: #else
1.146 noro 1649: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1650: #endif
1.1 noro 1651: if ( !stat ) {
1652: NEXT(l) = d; d = l;
1.103 noro 1653: d = nd_reconstruct(0,d);
1.1 noro 1654: goto again;
1655: } else if ( nf ) {
1.27 noro 1656: if ( checkonly ) return 0;
1.71 noro 1657: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1658: nd_removecont(m,nf);
1.118 noro 1659: if ( !m && nd_nalg ) {
1660: nd_monic(0,&nf);
1661: nd_removecont(m,nf);
1.117 noro 1662: }
1.63 noro 1663: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1664: nh = ndv_newps(m,nfv,0);
1.141 noro 1665: if ( !m && (ishomo && ++diag_count == diag_period) ) {
1.140 noro 1666: diag_count = 0;
1667: stat = do_diagonalize(sugar,m);
1668: if ( !stat ) {
1669: NEXT(l) = d; d = l;
1670: d = nd_reconstruct(1,d);
1671: goto again;
1672: }
1673: }
1.1 noro 1674: d = update_pairs(d,g,nh);
1675: g = update_base(g,nh);
1676: FREENDP(l);
1677: } else {
1.71 noro 1678: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1679: FREENDP(l);
1680: }
1681: }
1.77 noro 1682: if ( nd_demand )
1683: for ( t = g; t; t = NEXT(t) )
1684: BDY(t) = (pointer)ndv_load((int)BDY(t));
1685: else
1686: for ( t = g; t; t = NEXT(t) )
1687: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1688: return g;
1689: }
1690:
1.122 noro 1691: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1692: {
1693: int i,nh,stat;
1694: NODE r,g,t;
1695: ND h,nf,nfq,s,head;
1696: NDV nfv,nfqv;
1.113 noro 1697: Q q,den,num;
1.91 noro 1698:
1.143 noro 1699: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.91 noro 1700: /* for nd_ps */
1701: s = ndvtond(m,nd_ps[i]);
1702: s = nd_separate_head(s,&head);
1.122 noro 1703: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1704: if ( !stat ) return 0;
1.91 noro 1705: nf = nd_add(m,head,nf);
1706: ndv_free(nd_ps[i]);
1707: nd_ps[i] = ndtondv(m,nf);
1708: nd_free(nf);
1709:
1710: /* for nd_ps_trace */
1711: if ( nd_demand )
1712: nfv = ndv_load(i);
1713: else
1714: nfv = nd_ps_trace[i];
1715: s = ndvtond(0,nfv);
1716: s = nd_separate_head(s,&head);
1.146 noro 1717: stat = nd_nf(0,head,s,nd_ps_trace,1,0,&nf);
1.122 noro 1718: if ( !stat ) return 0;
1.91 noro 1719: ndv_free(nfv);
1720: nd_removecont(0,nf);
1721: nfv = ndtondv(0,nf);
1722: nd_free(nf);
1723: nd_bound[i] = ndv_compute_bound(nfv);
1724: if ( nd_demand ) {
1725: ndv_save(nfv,i);
1726: ndv_free(nfv);
1727: } else
1728: nd_ps_trace[i] = nfv;
1729: }
1.122 noro 1730: return 1;
1.91 noro 1731: }
1732:
1.118 noro 1733: static struct oEGT eg_invdalg;
1734: struct oEGT eg_le;
1735:
1.147 noro 1736: void nd_subst_vector(VL vl,P p,NODE subst,P *r)
1737: {
1738: NODE tn;
1739: P p1;
1740:
1741: for ( tn = subst; tn; tn = NEXT(NEXT(tn)) ) {
1742: substp(vl,p,BDY(tn),BDY(NEXT(tn)),&p1); p = p1;
1743: }
1744: *r = p;
1745: }
1746:
1.82 noro 1747: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1748: {
1749: int i,nh,sugar,stat;
1.23 noro 1750: NODE r,g,t;
1.20 noro 1751: ND_pairs d;
1752: ND_pairs l;
1.82 noro 1753: ND h,nf,nfq,s,head;
1.63 noro 1754: NDV nfv,nfqv;
1.113 noro 1755: Q q,den,num;
1.147 noro 1756: P hc;
1.113 noro 1757: union oNDC dn;
1.118 noro 1758: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1759: int diag_count = 0;
1.20 noro 1760:
1.118 noro 1761: init_eg(&eg_monic);
1762: init_eg(&eg_invdalg);
1763: init_eg(&eg_le);
1.23 noro 1764: g = 0; d = 0;
1765: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1766: d = update_pairs(d,g,i);
1767: g = update_base(g,i);
1768: }
1769: sugar = 0;
1770: while ( d ) {
1771: again:
1772: l = nd_minp(d,&d);
1773: if ( SG(l) != sugar ) {
1.130 noro 1774: #if 1
1.122 noro 1775: if ( ishomo ) {
1776: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1777: diag_count = 0;
1.122 noro 1778: if ( !stat ) {
1779: NEXT(l) = d; d = l;
1780: d = nd_reconstruct(1,d);
1781: goto again;
1782: }
1783: }
1.130 noro 1784: #endif
1.20 noro 1785: sugar = SG(l);
1.71 noro 1786: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1787: }
1.53 noro 1788: stat = nd_sp(m,0,l,&h);
1.20 noro 1789: if ( !stat ) {
1790: NEXT(l) = d; d = l;
1.103 noro 1791: d = nd_reconstruct(1,d);
1.20 noro 1792: goto again;
1793: }
1.41 noro 1794: #if USE_GEOBUCKET
1.53 noro 1795: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1796: #else
1.146 noro 1797: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1798: #endif
1.20 noro 1799: if ( !stat ) {
1800: NEXT(l) = d; d = l;
1.103 noro 1801: d = nd_reconstruct(1,d);
1.20 noro 1802: goto again;
1803: } else if ( nf ) {
1.77 noro 1804: if ( nd_demand ) {
1805: nfqv = ndv_load(nd_psn);
1806: nfq = ndvtond(0,nfqv);
1807: } else
1808: nfq = 0;
1809: if ( !nfq ) {
1.146 noro 1810: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,0,h,nd_ps_trace,!Top,0,&nfq) ) {
1.77 noro 1811: NEXT(l) = d; d = l;
1.103 noro 1812: d = nd_reconstruct(1,d);
1.77 noro 1813: goto again;
1814: }
1.70 noro 1815: }
1.20 noro 1816: if ( nfq ) {
1.70 noro 1817: /* m|HC(nfq) => failure */
1.147 noro 1818: if ( nd_vc ) {
1819: nd_subst_vector(nd_vc,HCP(nfq),nd_subst,&hc); q = (Q)hc;
1820: } else
1821: q = HCQ(nfq);
1822: if ( !rem(NM(q),m) ) return 0;
1.63 noro 1823:
1.71 noro 1824: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1825: if ( nd_nalg ) {
1826: /* m|DN(HC(nf)^(-1)) => failure */
1827: get_eg(&egm0);
1828: if ( !nd_monic(m,&nfq) ) return 0;
1829: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1830: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1831: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1832: } else {
1833: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1834: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1835: }
1.77 noro 1836: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1837: if ( ishomo && ++diag_count == diag_period ) {
1838: diag_count = 0;
1839: stat = do_diagonalize_trace(sugar,m);
1840: if ( !stat ) {
1841: NEXT(l) = d; d = l;
1842: d = nd_reconstruct(1,d);
1843: goto again;
1844: }
1845: }
1.20 noro 1846: d = update_pairs(d,g,nh);
1847: g = update_base(g,nh);
1848: } else {
1.71 noro 1849: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1850: }
1851: } else {
1.71 noro 1852: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1853: }
1854: FREENDP(l);
1855: }
1.77 noro 1856: if ( nd_demand )
1857: for ( t = g; t; t = NEXT(t) )
1858: BDY(t) = (pointer)ndv_load((int)BDY(t));
1859: else
1860: for ( t = g; t; t = NEXT(t) )
1861: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1862: if ( nd_nalg ) {
1863: print_eg("monic",&eg_monic);
1864: print_eg("invdalg",&eg_invdalg);
1865: print_eg("le",&eg_le);
1866: }
1.20 noro 1867: return g;
1868: }
1869:
1.23 noro 1870: int ndv_compare(NDV *p1,NDV *p2)
1871: {
1.34 noro 1872: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1873: }
1874:
1875: int ndv_compare_rev(NDV *p1,NDV *p2)
1876: {
1.34 noro 1877: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1878: }
1879:
1.61 noro 1880: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1881: {
1.69 noro 1882: int i,n,stat;
1883: ND nf,g,head;
1.23 noro 1884: NODE t,a0,a;
1.113 noro 1885: union oNDC dn;
1.69 noro 1886: NDV *w;
1.113 noro 1887: Q q,num,den;
1.23 noro 1888:
1.69 noro 1889: n = length(f);
1890: #if 0
1891: w = (NDV *)ALLOCA(n*sizeof(NDV));
1892: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1893: qsort(w,n,sizeof(NDV),
1894: (int (*)(const void *,const void *))ndv_compare);
1895: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1896: #endif
1.128 noro 1897: ndv_setup(m,0,f,0);
1.69 noro 1898: for ( i = 0; i < n; ) {
1899: g = ndvtond(m,nd_ps[i]);
1900: g = nd_separate_head(g,&head);
1.146 noro 1901: stat = nd_nf(m,head,g,nd_ps,1,0,&nf);
1.50 noro 1902: if ( !stat )
1.103 noro 1903: nd_reconstruct(0,0);
1.69 noro 1904: else {
1.71 noro 1905: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1906: ndv_free(nd_ps[i]);
1.24 noro 1907: nd_removecont(m,nf);
1.69 noro 1908: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1909: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1910: i++;
1.23 noro 1911: }
1912: }
1.71 noro 1913: if ( DP_Print ) { printf("\n"); }
1.23 noro 1914: for ( a0 = 0, i = 0; i < n; i++ ) {
1915: NEXTNODE(a0,a);
1.69 noro 1916: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1917: }
1918: NEXT(a) = 0;
1919: return a0;
1920: }
1921:
1.1 noro 1922: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1923: {
1924: ND_pairs d1,nd,cur,head,prev,remove;
1925:
1926: if ( !g ) return d;
1927: d = crit_B(d,t);
1928: d1 = nd_newpairs(g,t);
1929: d1 = crit_M(d1);
1930: d1 = crit_F(d1);
1.55 noro 1931: if ( do_weyl )
1932: head = d1;
1933: else {
1934: prev = 0; cur = head = d1;
1935: while ( cur ) {
1936: if ( crit_2( cur->i1,cur->i2 ) ) {
1937: remove = cur;
1938: if ( !prev ) head = cur = NEXT(cur);
1939: else cur = NEXT(prev) = NEXT(cur);
1940: FREENDP(remove);
1941: } else {
1942: prev = cur; cur = NEXT(cur);
1943: }
1.1 noro 1944: }
1945: }
1946: if ( !d )
1947: return head;
1948: else {
1949: nd = d;
1.34 noro 1950: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1951: NEXT(nd) = head;
1952: return d;
1953: }
1954: }
1955:
1956: ND_pairs nd_newpairs( NODE g, int t )
1957: {
1958: NODE h;
1.61 noro 1959: UINT *dl;
1.34 noro 1960: int ts,s;
1.1 noro 1961: ND_pairs r,r0;
1962:
1.20 noro 1963: dl = DL(nd_psh[t]);
1.34 noro 1964: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1965: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1966: NEXTND_pairs(r0,r);
1967: r->i1 = (int)BDY(h);
1968: r->i2 = t;
1.20 noro 1969: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1970: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1971: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1972: }
1973: NEXT(r) = 0;
1974: return r0;
1975: }
1976:
1977: ND_pairs crit_B( ND_pairs d, int s )
1978: {
1979: ND_pairs cur,head,prev,remove;
1.61 noro 1980: UINT *t,*tl,*lcm;
1.1 noro 1981: int td,tdl;
1982:
1983: if ( !d ) return 0;
1.20 noro 1984: t = DL(nd_psh[s]);
1.1 noro 1985: prev = 0;
1986: head = cur = d;
1.61 noro 1987: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1988: while ( cur ) {
1989: tl = cur->lcm;
1990: if ( ndl_reducible(tl,t)
1.20 noro 1991: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1992: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1993: remove = cur;
1994: if ( !prev ) {
1995: head = cur = NEXT(cur);
1996: } else {
1997: cur = NEXT(prev) = NEXT(cur);
1998: }
1999: FREENDP(remove);
2000: } else {
1.34 noro 2001: prev = cur; cur = NEXT(cur);
1.1 noro 2002: }
2003: }
2004: return head;
2005: }
2006:
2007: ND_pairs crit_M( ND_pairs d1 )
2008: {
2009: ND_pairs e,d2,d3,dd,p;
1.61 noro 2010: UINT *id,*jd;
1.1 noro 2011:
2012: for ( dd = 0, e = d1; e; e = d3 ) {
2013: if ( !(d2 = NEXT(e)) ) {
2014: NEXT(e) = dd;
2015: return e;
2016: }
1.34 noro 2017: id = LCM(e);
1.1 noro 2018: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2019: p = NEXT(d2);
2020: jd = LCM(d2);
2021: if ( ndl_equal(jd,id) )
2022: ;
2023: else if ( TD(jd) > TD(id) )
1.1 noro 2024: if ( ndl_reducible(jd,id) ) continue;
2025: else ;
1.34 noro 2026: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2027: NEXT(d2) = d3;
2028: d3 = d2;
2029: }
2030: NEXT(e) = dd;
2031: dd = e;
2032: continue;
2033: /**/
2034: delit: NEXT(d2) = d3;
2035: d3 = d2;
2036: for ( ; p; p = d2 ) {
2037: d2 = NEXT(p);
2038: NEXT(p) = d3;
2039: d3 = p;
2040: }
2041: FREENDP(e);
2042: }
2043: return dd;
2044: }
2045:
2046: ND_pairs crit_F( ND_pairs d1 )
2047: {
2048: ND_pairs rest, head,remove;
2049: ND_pairs last, p, r, w;
2050: int s;
2051:
2052: for ( head = last = 0, p = d1; NEXT(p); ) {
2053: r = w = equivalent_pairs(p,&rest);
1.14 noro 2054: s = SG(r);
1.1 noro 2055: w = NEXT(w);
2056: while ( w ) {
2057: if ( crit_2(w->i1,w->i2) ) {
2058: r = w;
2059: w = NEXT(w);
2060: while ( w ) {
2061: remove = w;
2062: w = NEXT(w);
2063: FREENDP(remove);
2064: }
2065: break;
1.14 noro 2066: } else if ( SG(w) < s ) {
1.1 noro 2067: FREENDP(r);
2068: r = w;
1.14 noro 2069: s = SG(r);
1.1 noro 2070: w = NEXT(w);
2071: } else {
2072: remove = w;
2073: w = NEXT(w);
2074: FREENDP(remove);
2075: }
2076: }
2077: if ( last ) NEXT(last) = r;
2078: else head = r;
2079: NEXT(last = r) = 0;
2080: p = rest;
2081: if ( !p ) return head;
2082: }
2083: if ( !last ) return p;
2084: NEXT(last) = p;
2085: return head;
2086: }
2087:
2088: int crit_2( int dp1, int dp2 )
2089: {
1.20 noro 2090: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2091: }
2092:
1.40 noro 2093: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2094: {
2095: ND_pairs w,p,r,s;
1.61 noro 2096: UINT *d;
1.1 noro 2097:
2098: w = d1;
1.34 noro 2099: d = LCM(w);
1.1 noro 2100: s = NEXT(w);
2101: NEXT(w) = 0;
2102: for ( r = 0; s; s = p ) {
2103: p = NEXT(s);
1.34 noro 2104: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2105: NEXT(s) = w; w = s;
1.1 noro 2106: } else {
1.39 noro 2107: NEXT(s) = r; r = s;
1.1 noro 2108: }
2109: }
2110: *prest = r;
2111: return w;
2112: }
2113:
2114: NODE update_base(NODE nd,int ndp)
2115: {
1.61 noro 2116: UINT *dl, *dln;
1.137 noro 2117: NODE last, p, head;
1.1 noro 2118:
1.20 noro 2119: dl = DL(nd_psh[ndp]);
1.1 noro 2120: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2121: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2122: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2123: p = NEXT(p);
2124: if ( last ) NEXT(last) = p;
2125: } else {
2126: if ( !last ) head = p;
2127: p = NEXT(last = p);
2128: }
2129: }
2130: head = append_one(head,ndp);
2131: return head;
2132: }
2133:
2134: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2135: {
2136: ND_pairs m,ml,p,l;
1.61 noro 2137: UINT *lcm;
1.33 noro 2138: int s,td,len,tlen,c,c1;
1.1 noro 2139:
2140: if ( !(p = NEXT(m = d)) ) {
2141: *prest = p;
2142: NEXT(m) = 0;
2143: return m;
2144: }
1.14 noro 2145: s = SG(m);
1.77 noro 2146: if ( !NoSugar ) {
2147: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2148: if ( (SG(p) < s)
2149: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2150: ml = l; m = p; s = SG(m);
2151: }
2152: } else {
2153: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2154: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2155: ml = l; m = p; s = SG(m);
2156: }
2157: }
1.1 noro 2158: if ( !ml ) *prest = NEXT(m);
2159: else {
2160: NEXT(ml) = NEXT(m);
2161: *prest = d;
2162: }
2163: NEXT(m) = 0;
2164: return m;
2165: }
2166:
1.63 noro 2167: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2168: {
1.69 noro 2169: int msugar,i;
1.63 noro 2170: ND_pairs t,dm0,dm,dr0,dr;
2171:
2172: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2173: if ( SG(t) < msugar ) msugar = SG(t);
2174: dm0 = 0; dr0 = 0;
1.69 noro 2175: for ( i = 0, t = d; t; t = NEXT(t) )
2176: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2177: if ( dm0 ) NEXT(dm) = t;
2178: else dm0 = t;
2179: dm = t;
1.69 noro 2180: i++;
1.63 noro 2181: } else {
2182: if ( dr0 ) NEXT(dr) = t;
2183: else dr0 = t;
2184: dr = t;
2185: }
2186: NEXT(dm) = 0;
2187: if ( dr0 ) NEXT(dr) = 0;
2188: *prest = dr0;
2189: return dm0;
2190: }
2191:
1.77 noro 2192: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2193: {
1.3 noro 2194: int len;
1.13 noro 2195: RHist r;
1.20 noro 2196: NDV b;
1.3 noro 2197:
1.1 noro 2198: if ( nd_psn == nd_pslen ) {
2199: nd_pslen *= 2;
1.11 noro 2200: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2201: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2202: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2203: nd_bound = (UINT **)
2204: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2205: }
1.39 noro 2206: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2207: nd_ps[nd_psn] = a;
1.39 noro 2208: if ( aq ) {
1.63 noro 2209: nd_ps_trace[nd_psn] = aq;
2210: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2211: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2212: } else {
1.63 noro 2213: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2214: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2215: }
1.77 noro 2216: if ( nd_demand ) {
2217: if ( aq ) {
2218: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2219: nd_ps_trace[nd_psn] = 0;
2220: } else {
2221: ndv_save(nd_ps[nd_psn],nd_psn);
2222: nd_ps[nd_psn] = 0;
2223: }
2224: }
1.1 noro 2225: return nd_psn++;
2226: }
2227:
1.128 noro 2228: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2229: {
1.5 noro 2230: int i,j,td,len,max;
1.1 noro 2231: NODE s,s0,f0;
1.61 noro 2232: UINT *d;
1.13 noro 2233: RHist r;
1.69 noro 2234: NDV *w;
1.61 noro 2235: NDV a,am;
1.11 noro 2236:
2237: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2238:
1.69 noro 2239: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2240: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2241: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.139 noro 2242: if ( !dont_sort ) {
2243: /* XXX heuristic */
2244: if ( !nd_ord->id && (nd_ord->ord.simple<2) )
2245: qsort(w,nd_psn,sizeof(NDV),
2246: (int (*)(const void *,const void *))ndv_compare_rev);
2247: else
2248: qsort(w,nd_psn,sizeof(NDV),
2249: (int (*)(const void *,const void *))ndv_compare);
2250: }
1.69 noro 2251: nd_pslen = 2*nd_psn;
1.11 noro 2252: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2253: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2254: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2255: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2256:
1.146 noro 2257: if ( trace && nd_vc )
2258: makesubst(nd_vc,&nd_subst);
2259: else
2260: nd_subst = 0;
2261:
1.11 noro 2262: if ( !nd_red )
1.13 noro 2263: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2264: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2265: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2266: if ( trace ) {
1.69 noro 2267: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2268: ndv_removecont(0,a);
2269: am = nd_ps[i] = ndv_dup(mod,a);
2270: ndv_mod(mod,am);
2271: ndv_removecont(mod,am);
2272: } else {
1.69 noro 2273: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2274: ndv_removecont(mod,a);
1.39 noro 2275: }
1.61 noro 2276: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2277: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2278: nd_psh[i] = r;
1.77 noro 2279: if ( nd_demand ) {
2280: if ( trace ) {
2281: ndv_save(nd_ps_trace[i],i);
2282: nd_ps_trace[i] = 0;
2283: } else {
2284: ndv_save(nd_ps[i],i);
2285: nd_ps[i] = 0;
2286: }
2287: }
1.20 noro 2288: }
2289: }
2290:
1.119 noro 2291: struct order_spec *append_block(struct order_spec *spec,
2292: int nv,int nalg,int ord);
2293:
1.121 noro 2294: extern VECT current_dl_weight_vector_obj;
2295: static VECT prev_weight_vector_obj;
2296:
1.120 noro 2297: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2298: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2299: {
1.121 noro 2300: NODE alist,t,s,r0,r,arg;
1.120 noro 2301: VL tv;
2302: P poly;
2303: DP d;
2304: Alg alpha,dp;
2305: DAlg inv,da,hc;
2306: MP m;
1.121 noro 2307: int i,nvar,nalg,n;
1.120 noro 2308: NumberField nf;
2309: LIST f1,f2;
2310: struct order_spec *current_spec;
1.121 noro 2311: VECT obj,obj0;
2312: Obj tmp;
1.120 noro 2313:
2314: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2315: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2316:
2317: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2318: NEXTNODE(alist,t); MKV(tv->v,poly);
2319: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2320: tv->v = tv->v->priv;
2321: }
2322: NEXT(t) = 0;
2323:
2324: /* simplification, makeing polynomials monic */
2325: setfield_dalg(alist);
2326: obj_algtodalg(f,&f1);
2327: for ( t = BDY(f); t; t = NEXT(t) ) {
2328: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2329: hc = (DAlg)BDY(d)->c;
2330: if ( NID(hc) == N_DA ) {
2331: invdalg(hc,&inv);
2332: for ( m = BDY(d); m; m = NEXT(m) ) {
2333: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2334: }
2335: }
2336: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2337: }
2338: obj_dalgtoalg(f1,&f);
2339:
2340: /* append alg vars to the var list */
2341: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2342: NEXT(tv) = av;
2343:
2344: /* append a block to ord */
2345: *ord1p = append_block(ord,nvar,nalg,2);
2346:
2347: /* create generator list */
2348: nf = get_numberfield();
2349: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2350: MKAlg(nf->defpoly[i],dp);
2351: MKNODE(s,dp,t); t = s;
2352: }
2353: MKLIST(f1,t);
2354: *alistp = alist;
2355: algobjtorat(f1,f1p);
1.121 noro 2356:
2357: /* creating a new weight vector */
2358: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2359: n = nvar+nalg+1;
2360: MKVECT(obj,n);
2361: if ( obj0 && obj0->len == nvar )
2362: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2363: else
2364: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2365: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2366: BDY(obj)[n-1] = (pointer)ONE;
2367: arg = mknode(1,obj);
2368: Pdp_set_weight(arg,&tmp);
2369: }
2370:
2371: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2372: {
2373: NODE s,t,u0,u;
2374: P p;
2375: VL tv;
2376: Obj obj,tmp;
2377: NODE arg;
2378:
2379: u0 = 0;
2380: for ( t = r; t; t = NEXT(t) ) {
2381: p = (P)BDY(t);
2382: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2383: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2384: }
2385: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2386: NEXTNODE(u0,u);
2387: BDY(u) = (pointer)p;
2388: }
2389: }
2390: arg = mknode(1,prev_weight_vector_obj);
2391: Pdp_set_weight(arg,&tmp);
2392:
2393: return u0;
1.120 noro 2394: }
2395:
1.63 noro 2396: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2397: {
1.119 noro 2398: VL tv,fv,vv,vc,av;
2399: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2400: int e,max,nvar,i;
1.69 noro 2401: NDV b;
1.119 noro 2402: int ishomo,nalg;
2403: Alg alpha,dp;
1.144 noro 2404: P p,zp;
2405: Q dmy;
1.119 noro 2406: LIST f1,f2;
2407: Obj obj;
2408: NumberField nf;
1.120 noro 2409: struct order_spec *ord1;
1.1 noro 2410:
1.78 noro 2411: if ( !m && Demand ) nd_demand = 1;
2412: else nd_demand = 0;
2413:
1.103 noro 2414: #if 0
1.74 noro 2415: ndv_alloc = 0;
1.103 noro 2416: #endif
1.146 noro 2417: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.74 noro 2418: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2419: switch ( ord->id ) {
2420: case 1:
2421: if ( ord->nv != nvar )
2422: error("nd_{gr,f4} : invalid order specification");
2423: break;
2424: default:
2425: break;
2426: }
1.119 noro 2427: nd_nalg = 0;
1.123 noro 2428: av = 0;
1.119 noro 2429: if ( !m ) {
2430: get_algtree((Obj)f,&av);
2431: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2432: nd_ntrans = nvar;
1.119 noro 2433: nd_nalg = nalg;
2434: /* #i -> t#i */
2435: if ( nalg ) {
1.120 noro 2436: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2437: ord = ord1;
2438: f = f1;
1.119 noro 2439: }
1.120 noro 2440: nvar += nalg;
1.119 noro 2441: }
1.32 noro 2442: nd_init_ord(ord);
1.61 noro 2443: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2444: for ( tv = vv; tv; tv = NEXT(tv) ) {
2445: e = getdeg(tv->v,(P)BDY(t));
2446: max = MAX(e,max);
2447: }
2448: nd_setup_parameters(nvar,max);
1.77 noro 2449: ishomo = 1;
1.1 noro 2450: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2451: ptozp((P)BDY(t),1,&dmy,&zp);
2452: b = (pointer)ptondv(CO,vv,zp);
1.77 noro 2453: if ( ishomo )
2454: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2455: if ( m ) ndv_mod(m,b);
1.61 noro 2456: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2457: }
2458: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2459: ndv_setup(m,0,fd0,0);
1.77 noro 2460: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2461: nd_demand = 0;
1.61 noro 2462: x = ndv_reducebase(x);
2463: x = ndv_reduceall(m,x);
1.23 noro 2464: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2465: NEXTNODE(r0,r);
1.61 noro 2466: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2467: }
2468: if ( r0 ) NEXT(r) = 0;
1.121 noro 2469: if ( nalg )
2470: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2471: MKLIST(*rp,r0);
1.103 noro 2472: #if 0
1.74 noro 2473: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2474: #endif
1.127 noro 2475: }
2476:
2477: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2478: {
2479: VL tv,fv,vv,vc,av;
2480: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2481: int e,max,nvar,i;
2482: NDV b;
2483: int ishomo,nalg;
2484: Alg alpha,dp;
1.144 noro 2485: P p,zp;
2486: Q dmy;
1.127 noro 2487: LIST f1,f2;
2488: Obj obj;
2489: NumberField nf;
2490: struct order_spec *ord1;
2491:
1.146 noro 2492: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.127 noro 2493: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2494: switch ( ord->id ) {
2495: case 1:
2496: if ( ord->nv != nvar )
2497: error("nd_check : invalid order specification");
2498: break;
2499: default:
2500: break;
2501: }
2502: nd_nalg = 0;
2503: av = 0;
2504: if ( !m ) {
2505: get_algtree((Obj)f,&av);
2506: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2507: nd_ntrans = nvar;
2508: nd_nalg = nalg;
2509: /* #i -> t#i */
2510: if ( nalg ) {
2511: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2512: ord = ord1;
2513: f = f1;
2514: }
2515: nvar += nalg;
2516: }
2517: nd_init_ord(ord);
2518: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2519: for ( tv = vv; tv; tv = NEXT(tv) ) {
2520: e = getdeg(tv->v,(P)BDY(t));
2521: max = MAX(e,max);
2522: }
2523: nd_setup_parameters(nvar,max);
2524: ishomo = 1;
2525: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2526: ptozp((P)BDY(t),1,&dmy,&zp);
2527: b = (pointer)ptondv(CO,vv,zp);
1.127 noro 2528: if ( ishomo )
2529: ishomo = ishomo && ndv_ishomo(b);
2530: if ( m ) ndv_mod(m,b);
2531: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2532: }
2533: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2534: ndv_setup(m,0,fd0,0);
1.127 noro 2535: for ( x = 0, i = 0; i < nd_psn; i++ )
2536: x = update_base(x,i);
2537: if ( do_check ) {
2538: x = nd_gb(m,ishomo,1);
2539: if ( !x ) {
2540: *rp = 0;
2541: return;
2542: }
2543: } else {
2544: for ( t = x; t; t = NEXT(t) )
2545: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2546: }
2547: x = ndv_reducebase(x);
2548: x = ndv_reduceall(m,x);
2549: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2550: NEXTNODE(r0,r);
2551: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2552: }
2553: if ( r0 ) NEXT(r) = 0;
2554: if ( nalg )
2555: r0 = postprocess_algcoef(av,alist,r0);
2556: MKLIST(*rp,r0);
1.20 noro 2557: }
2558:
1.133 noro 2559: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2560: {
1.119 noro 2561: VL tv,fv,vv,vc,av;
2562: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2563: int m,nocheck,nvar,mindex,e,max;
2564: NDV c;
2565: NMV a;
1.144 noro 2566: P p,zp;
2567: Q dmy;
1.61 noro 2568: EPOS oepos;
1.119 noro 2569: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2570: Alg alpha,dp;
2571: P poly;
2572: LIST f1,f2;
2573: Obj obj;
2574: NumberField nf;
1.120 noro 2575: struct order_spec *ord1;
1.145 noro 2576: struct oEGT eg_check,eg0,eg1;
1.20 noro 2577:
1.146 noro 2578: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.74 noro 2579: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2580: switch ( ord->id ) {
2581: case 1:
2582: if ( ord->nv != nvar )
2583: error("nd_gr_trace : invalid order specification");
2584: break;
2585: default:
2586: break;
2587: }
1.119 noro 2588:
2589: get_algtree((Obj)f,&av);
2590: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2591: nd_ntrans = nvar;
1.119 noro 2592: nd_nalg = nalg;
2593: /* #i -> t#i */
2594: if ( nalg ) {
1.120 noro 2595: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2596: ord = ord1;
2597: f = f1;
1.119 noro 2598: }
1.120 noro 2599: nvar += nalg;
1.119 noro 2600:
1.52 noro 2601: nocheck = 0;
2602: mindex = 0;
2603:
1.78 noro 2604: if ( Demand ) nd_demand = 1;
2605: else nd_demand = 0;
2606:
1.52 noro 2607: /* setup modulus */
2608: if ( trace < 0 ) {
2609: trace = -trace;
2610: nocheck = 1;
2611: }
2612: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2613: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2614: for ( tv = vv; tv; tv = NEXT(tv) ) {
2615: e = getdeg(tv->v,(P)BDY(t));
2616: max = MAX(e,max);
1.23 noro 2617: }
1.61 noro 2618: nd_init_ord(ord);
2619: nd_setup_parameters(nvar,max);
2620: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2621: ishomo = 1;
1.61 noro 2622: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2623: ptozp((P)BDY(t),1,&dmy,&zp);
2624: c = (pointer)ptondv(CO,vv,zp);
1.82 noro 2625: if ( ishomo )
2626: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2627: if ( c ) {
2628: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2629: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2630: }
1.61 noro 2631: }
2632: if ( in0 ) NEXT(in) = 0;
2633: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2634: if ( !ishomo && homo ) {
1.122 noro 2635: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2636: c = (NDV)BDY(t); len = LEN(c);
2637: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2638: wmax = MAX(TD(DL(a)),wmax);
2639: }
2640: homogenize_order(ord,nvar,&ord1);
1.89 noro 2641: nd_init_ord(ord1);
1.61 noro 2642: nd_setup_parameters(nvar+1,wmax);
2643: for ( t = fd0; t; t = NEXT(t) )
2644: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2645: }
1.52 noro 2646: while ( 1 ) {
1.77 noro 2647: if ( Demand )
2648: nd_demand = 1;
1.128 noro 2649: ndv_setup(m,1,fd0,0);
1.133 noro 2650: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
1.52 noro 2651: if ( !cand ) {
2652: /* failure */
1.61 noro 2653: if ( trace > 1 ) { *rp = 0; return; }
2654: else m = get_lprime(++mindex);
1.52 noro 2655: continue;
2656: }
1.82 noro 2657: if ( !ishomo && homo ) {
1.27 noro 2658: /* dehomogenization */
1.61 noro 2659: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2660: nd_init_ord(ord);
1.61 noro 2661: nd_setup_parameters(nvar,0);
1.27 noro 2662: }
1.77 noro 2663: nd_demand = 0;
1.61 noro 2664: cand = ndv_reducebase(cand);
2665: cand = ndv_reduceall(0,cand);
1.78 noro 2666: cbpe = nd_bpe;
1.61 noro 2667: if ( nocheck )
2668: break;
1.145 noro 2669: get_eg(&eg0);
1.61 noro 2670: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2671: /* success */
2672: break;
2673: else if ( trace > 1 ) {
2674: /* failure */
2675: *rp = 0; return;
1.61 noro 2676: } else {
1.52 noro 2677: /* try the next modulus */
2678: m = get_lprime(++mindex);
1.61 noro 2679: /* reset the parameters */
1.82 noro 2680: if ( !ishomo && homo ) {
1.89 noro 2681: nd_init_ord(ord1);
1.61 noro 2682: nd_setup_parameters(nvar+1,wmax);
2683: } else {
2684: nd_init_ord(ord);
2685: nd_setup_parameters(nvar,max);
2686: }
2687: }
1.52 noro 2688: }
1.145 noro 2689: get_eg(&eg1); init_eg(&eg_check); add_eg(&eg_check,&eg0,&eg1);
2690: if ( DP_Print )
2691: fprintf(asir_out,"check=%fsec\n",eg_check.exectime+eg_check.gctime);
1.27 noro 2692: /* dp->p */
1.61 noro 2693: nd_bpe = cbpe;
1.73 noro 2694: nd_setup_parameters(nd_nvar,0);
1.121 noro 2695: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2696: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2697: if ( nalg )
2698: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2699: MKLIST(*rp,cand);
1.1 noro 2700: }
2701:
1.61 noro 2702: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2703: {
1.61 noro 2704: UINT *d;
1.57 noro 2705: int i,j,l,s,ord_l;
1.43 noro 2706: struct order_pair *op;
1.1 noro 2707:
2708: d = dl->d;
1.41 noro 2709: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2710: if ( nd_blockmask ) {
2711: l = nd_blockmask->n;
2712: op = nd_blockmask->order_pair;
2713: for ( j = 0, s = 0; j < l; j++ ) {
2714: ord_l = op[j].length;
1.57 noro 2715: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2716: }
2717: TD(r) = ndl_weight(r);
1.61 noro 2718: ndl_weight_mask(r);
1.43 noro 2719: } else {
1.56 noro 2720: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2721: TD(r) = ndl_weight(r);
2722: }
1.1 noro 2723: }
2724:
1.61 noro 2725: DL ndltodl(int n,UINT *ndl)
1.1 noro 2726: {
2727: DL dl;
2728: int *d;
1.57 noro 2729: int i,j,l,s,ord_l;
1.43 noro 2730: struct order_pair *op;
1.1 noro 2731:
2732: NEWDL(dl,n);
1.34 noro 2733: dl->td = TD(ndl);
1.1 noro 2734: d = dl->d;
1.43 noro 2735: if ( nd_blockmask ) {
2736: l = nd_blockmask->n;
2737: op = nd_blockmask->order_pair;
2738: for ( j = 0, s = 0; j < l; j++ ) {
2739: ord_l = op[j].length;
1.57 noro 2740: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2741: }
2742: } else {
1.56 noro 2743: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2744: }
1.1 noro 2745: return dl;
2746: }
2747:
1.61 noro 2748: void ndl_print(UINT *dl)
1.1 noro 2749: {
2750: int n;
1.57 noro 2751: int i,j,l,ord_l,s,s0;
1.43 noro 2752: struct order_pair *op;
1.1 noro 2753:
2754: n = nd_nvar;
2755: printf("<<");
1.43 noro 2756: if ( nd_blockmask ) {
2757: l = nd_blockmask->n;
2758: op = nd_blockmask->order_pair;
2759: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2760: ord_l = op[j].length;
1.57 noro 2761: for ( i = 0; i < ord_l; i++, s++ )
2762: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2763: }
2764: } else {
1.56 noro 2765: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2766: }
1.1 noro 2767: printf(">>");
2768: }
2769:
2770: void nd_print(ND p)
2771: {
2772: NM m;
2773:
2774: if ( !p )
2775: printf("0\n");
2776: else {
2777: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2778: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2779: else printf("+%d*",CM(m));
1.14 noro 2780: ndl_print(DL(m));
1.1 noro 2781: }
2782: printf("\n");
2783: }
2784: }
2785:
1.113 noro 2786: void nd_print_q(ND p)
1.16 noro 2787: {
2788: NM m;
2789:
2790: if ( !p )
2791: printf("0\n");
2792: else {
2793: for ( m = BDY(p); m; m = NEXT(m) ) {
2794: printf("+");
1.113 noro 2795: printexpr(CO,(Obj)CQ(m));
1.16 noro 2796: printf("*");
2797: ndl_print(DL(m));
2798: }
2799: printf("\n");
2800: }
2801: }
2802:
1.1 noro 2803: void ndp_print(ND_pairs d)
2804: {
2805: ND_pairs t;
2806:
1.34 noro 2807: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2808: printf("\n");
2809: }
2810:
1.20 noro 2811: void nd_removecont(int mod,ND p)
1.16 noro 2812: {
2813: int i,n;
1.113 noro 2814: Q *w;
2815: Q dvr,t;
1.16 noro 2816: NM m;
1.21 noro 2817: struct oVECT v;
1.113 noro 2818: N q,r;
1.16 noro 2819:
1.71 noro 2820: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2821: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2822: else {
2823: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2824: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2825: v.len = n;
2826: v.body = (pointer *)w;
1.113 noro 2827: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.146 noro 2828: removecont_array((P *)w,n);
1.113 noro 2829: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2830: }
2831: }
2832:
1.21 noro 2833: void nd_removecont2(ND p1,ND p2)
2834: {
2835: int i,n1,n2,n;
1.113 noro 2836: Q *w;
2837: Q dvr,t;
1.21 noro 2838: NM m;
2839: struct oVECT v;
1.113 noro 2840: N q,r;
1.21 noro 2841:
2842: if ( !p1 ) {
2843: nd_removecont(0,p2); return;
2844: } else if ( !p2 ) {
2845: nd_removecont(0,p1); return;
2846: }
2847: n1 = nd_length(p1);
2848: n2 = nd_length(p2);
2849: n = n1+n2;
1.113 noro 2850: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2851: v.len = n;
2852: v.body = (pointer *)w;
1.113 noro 2853: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2854: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.146 noro 2855: removecont_array((P *)w,n);
1.113 noro 2856: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2857: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2858: }
2859:
1.20 noro 2860: void ndv_removecont(int mod,NDV p)
1.16 noro 2861: {
1.146 noro 2862: int i,len,all_p;
2863: Q *c;
2864: P *w;
1.113 noro 2865: Q dvr,t;
1.146 noro 2866: P g,cont,tp;
1.16 noro 2867: NMV m;
2868:
1.71 noro 2869: if ( mod == -1 )
2870: ndv_mul_c(mod,p,_invsf(HCM(p)));
2871: else if ( mod )
1.20 noro 2872: ndv_mul_c(mod,p,invm(HCM(p),mod));
2873: else {
2874: len = p->len;
1.146 noro 2875: w = (P *)ALLOCA(len*sizeof(P));
2876: c = (Q *)ALLOCA(len*sizeof(Q));
2877: for ( m = BDY(p), all_p = 1, i = 0; i < len; NMV_ADV(m), i++ ) {
2878: ptozp(CP(m),1,&c[i],&w[i]);
2879: all_p = all_p && !NUM(w[i]);
2880: }
2881: if ( all_p ) {
2882: qltozl(c,len,&dvr); heu_nezgcdnpz(nd_vc,w,len,&g);
2883: mulp(nd_vc,(P)dvr,g,&cont);
2884: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2885: divsp(nd_vc,CP(m),cont,&tp); CP(m) = tp;
2886: }
2887: } else {
2888: sortbynm((Q *)c,len);
2889: qltozl((Q *)c,len,&dvr);
2890: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2891: divsp(nd_vc,CP(m),(P)dvr,&tp); CP(m) = tp;
2892: }
1.113 noro 2893: }
1.16 noro 2894: }
1.21 noro 2895: }
2896:
1.61 noro 2897: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2898: {
2899: int len,i,max;
2900: NMV m,mr0,mr,t;
2901:
2902: len = p->len;
2903: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2904: max = MAX(max,TD(DL(m)));
2905: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2906: m = (NMV)((char *)mr0+(len-1)*oadv);
2907: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2908: t = (NMV)ALLOCA(nmv_adv);
2909: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2910: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2911: CQ(mr) = CQ(m);
1.61 noro 2912: ndl_copy(DL(t),DL(mr));
2913: }
2914: NV(p)++;
2915: BDY(p) = mr0;
2916: }
2917:
1.45 noro 2918: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2919: {
1.45 noro 2920: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2921: Q *w;
2922: Q dvr,t;
1.23 noro 2923: NMV m,r;
2924:
2925: len = p->len;
2926: newnvar = nd_nvar-1;
1.48 noro 2927: newexporigin = nd_get_exporigin(ord);
1.45 noro 2928: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2929: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2930: ndl_dehomogenize(DL(m));
1.23 noro 2931: if ( newwpd != nd_wpd ) {
1.90 noro 2932: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2933: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2934: CQ(r) = CQ(m);
1.45 noro 2935: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2936: adj = nd_exporigin-newexporigin;
2937: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2938: }
2939: }
2940: NV(p)--;
2941: }
2942:
1.146 noro 2943: void removecont_array(P *p,int n)
2944: {
2945: int all_p,all_q,i;
2946: Q *c;
2947: P *w;
2948: P t,s;
2949:
2950: for ( all_q = 1, i = 0; i < n; i++ )
2951: all_q = all_q && NUM(p[i]);
2952: if ( all_q ) {
2953: removecont_array_q((Q *)p,n);
2954: } else {
2955: c = (Q *)ALLOCA(n*sizeof(Q));
2956: w = (P *)ALLOCA(n*sizeof(P));
2957: for ( i = 0; i < n; i++ ) {
2958: ptozp(p[i],1,&c[i],&w[i]);
2959: }
2960: removecont_array_q(c,n);
2961: heu_nezgcdnpz(nd_vc,w,n,&t);
2962: for ( i = 0; i < n; i++ ) {
2963: divsp(nd_vc,w[i],t,&s); mulp(nd_vc,s,(P)c[i],&p[i]);
2964: }
2965: }
2966: }
2967:
2968: void removecont_array_q(Q *c,int n)
1.21 noro 2969: {
2970: struct oVECT v;
1.113 noro 2971: Q d0,d1,a,u,u1,gcd;
1.110 noro 2972: int i,j;
1.113 noro 2973: N qn,rn,gn;
2974: Q *q,*r;
1.21 noro 2975:
1.113 noro 2976: q = (Q *)ALLOCA(n*sizeof(Q));
2977: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2978: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2979: igcdv_estimate(&v,&d0);
2980: for ( i = 0; i < n; i++ ) {
2981: divn(NM(c[i]),NM(d0),&qn,&rn);
2982: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2983: NTOQ(rn,SGN(c[i]),r[i]);
2984: }
1.34 noro 2985: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2986: if ( i < n ) {
2987: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2988: igcdv(&v,&d1);
2989: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2990: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 2991: for ( i = 0; i < n; i++ ) {
1.113 noro 2992: mulq(a,q[i],&u);
2993: if ( r[i] ) {
2994: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2995: addq(u,u1,&q[i]);
2996: } else
2997: q[i] = u;
1.21 noro 2998: }
2999: }
1.34 noro 3000: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 3001: }
3002:
1.19 noro 3003: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 3004: {
3005: NM m;
3006: int c,c1;
3007:
1.34 noro 3008: if ( !p ) return;
1.115 noro 3009: if ( mul == 1 ) return;
1.71 noro 3010: if ( mod == -1 )
3011: for ( m = BDY(p); m; m = NEXT(m) )
3012: CM(m) = _mulsf(CM(m),mul);
3013: else
3014: for ( m = BDY(p); m; m = NEXT(m) ) {
3015: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3016: }
1.1 noro 3017: }
3018:
1.146 noro 3019: void nd_mul_c_q(ND p,P mul)
1.16 noro 3020: {
3021: NM m;
1.146 noro 3022: P c;
1.16 noro 3023:
1.113 noro 3024: if ( !p ) return;
1.115 noro 3025: if ( UNIQ(mul) ) return;
1.16 noro 3026: for ( m = BDY(p); m; m = NEXT(m) ) {
1.146 noro 3027: mulp(nd_vc,CP(m),mul,&c); CP(m) = c;
1.16 noro 3028: }
3029: }
3030:
1.61 noro 3031: void nd_mul_c_p(VL vl,ND p,P mul)
3032: {
3033: NM m;
3034: P c;
3035:
3036: if ( !p ) return;
3037: for ( m = BDY(p); m; m = NEXT(m) ) {
3038: mulp(vl,CP(m),mul,&c); CP(m) = c;
3039: }
3040: }
3041:
1.1 noro 3042: void nd_free(ND p)
3043: {
3044: NM t,s;
3045:
1.34 noro 3046: if ( !p ) return;
1.1 noro 3047: t = BDY(p);
3048: while ( t ) {
3049: s = NEXT(t);
3050: FREENM(t);
3051: t = s;
3052: }
3053: FREEND(p);
3054: }
3055:
1.23 noro 3056: void ndv_free(NDV p)
3057: {
3058: GC_free(BDY(p));
3059: }
3060:
1.61 noro 3061: void nd_append_red(UINT *d,int i)
1.1 noro 3062: {
1.13 noro 3063: RHist m,m0;
1.1 noro 3064: int h;
3065:
1.13 noro 3066: NEWRHist(m);
1.34 noro 3067: h = ndl_hash_value(d);
1.13 noro 3068: m->index = i;
1.14 noro 3069: ndl_copy(d,DL(m));
1.1 noro 3070: NEXT(m) = nd_red[h];
3071: nd_red[h] = m;
3072: }
3073:
1.61 noro 3074: UINT *ndv_compute_bound(NDV p)
1.1 noro 3075: {
1.61 noro 3076: UINT *d1,*d2,*t;
3077: UINT u;
1.57 noro 3078: int i,j,k,l,len,ind;
1.45 noro 3079: NMV m;
1.1 noro 3080:
3081: if ( !p )
3082: return 0;
1.61 noro 3083: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3084: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 3085: len = LEN(p);
3086: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3087: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3088: ndl_lcm(DL(m),d1,d2);
1.1 noro 3089: t = d1; d1 = d2; d2 = t;
3090: }
1.12 noro 3091: l = nd_nvar+31;
1.61 noro 3092: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3093: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3094: u = d1[i];
3095: k = (nd_epw-1)*nd_bpe;
3096: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3097: t[ind] = (u>>k)&nd_mask0;
3098: }
3099: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3100: return t;
3101: }
3102:
1.99 noro 3103: UINT *nd_compute_bound(ND p)
3104: {
3105: UINT *d1,*d2,*t;
3106: UINT u;
3107: int i,j,k,l,len,ind;
3108: NM m;
3109:
3110: if ( !p )
3111: return 0;
3112: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3113: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3114: len = LEN(p);
3115: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3116: for ( m = NEXT(m); m; m = NEXT(m) ) {
3117: ndl_lcm(DL(m),d1,d2);
3118: t = d1; d1 = d2; d2 = t;
3119: }
3120: l = nd_nvar+31;
3121: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3122: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3123: u = d1[i];
3124: k = (nd_epw-1)*nd_bpe;
3125: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3126: t[ind] = (u>>k)&nd_mask0;
3127: }
3128: for ( ; ind < l; ind++ ) t[ind] = 0;
3129: return t;
3130: }
3131:
1.48 noro 3132: int nd_get_exporigin(struct order_spec *ord)
3133: {
1.51 noro 3134: switch ( ord->id ) {
1.96 noro 3135: case 0: case 2:
1.48 noro 3136: return 1;
1.41 noro 3137: case 1:
3138: /* block order */
1.43 noro 3139: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3140: return ord->ord.block.length+1;
1.96 noro 3141: case 3:
3142: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3143: }
1.48 noro 3144: }
3145:
1.61 noro 3146: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3147: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3148: struct order_pair *op;
1.48 noro 3149:
1.73 noro 3150: nd_nvar = nvar;
1.79 noro 3151: if ( max ) {
1.83 noro 3152: /* XXX */
3153: if ( do_weyl ) nd_bpe = 32;
3154: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3155: else if ( max < 4 ) nd_bpe = 2;
3156: else if ( max < 8 ) nd_bpe = 3;
3157: else if ( max < 16 ) nd_bpe = 4;
3158: else if ( max < 32 ) nd_bpe = 5;
3159: else if ( max < 64 ) nd_bpe = 6;
3160: else if ( max < 256 ) nd_bpe = 8;
3161: else if ( max < 1024 ) nd_bpe = 10;
3162: else if ( max < 65536 ) nd_bpe = 16;
3163: else nd_bpe = 32;
3164: }
1.61 noro 3165: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3166: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3167: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3168: wpd = nd_exporigin+elen;
3169: if ( wpd != nd_wpd ) {
3170: nd_free_private_storage();
3171: nd_wpd = wpd;
3172: }
1.1 noro 3173: if ( nd_bpe < 32 ) {
3174: nd_mask0 = (1<<nd_bpe)-1;
3175: } else {
3176: nd_mask0 = 0xffffffff;
3177: }
3178: bzero(nd_mask,sizeof(nd_mask));
3179: nd_mask1 = 0;
3180: for ( i = 0; i < nd_epw; i++ ) {
3181: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3182: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3183: }
1.90 noro 3184: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3185: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3186: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3187: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3188: }
3189:
1.103 noro 3190: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3191: {
1.37 noro 3192: int i,obpe,oadv,h;
1.71 noro 3193: static NM prev_nm_free_list;
3194: static ND_pairs prev_ndp_free_list;
1.13 noro 3195: RHist mr0,mr;
3196: RHist r;
1.37 noro 3197: RHist *old_red;
1.71 noro 3198: ND_pairs s0,s,t;
1.43 noro 3199: EPOS oepos;
1.15 noro 3200:
1.1 noro 3201: obpe = nd_bpe;
1.11 noro 3202: oadv = nmv_adv;
1.43 noro 3203: oepos = nd_epos;
1.63 noro 3204: if ( obpe < 2 ) nd_bpe = 2;
3205: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3206: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3207: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3208: else if ( obpe < 6 ) nd_bpe = 6;
3209: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3210: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3211: else if ( obpe < 16 ) nd_bpe = 16;
3212: else if ( obpe < 32 ) nd_bpe = 32;
3213: else error("nd_reconstruct : exponent too large");
1.5 noro 3214:
1.73 noro 3215: nd_setup_parameters(nd_nvar,0);
1.1 noro 3216: prev_nm_free_list = _nm_free_list;
3217: prev_ndp_free_list = _ndp_free_list;
3218: _nm_free_list = 0;
3219: _ndp_free_list = 0;
1.53 noro 3220: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3221: if ( trace )
3222: for ( i = nd_psn-1; i >= 0; i-- )
3223: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3224: s0 = 0;
3225: for ( t = d; t; t = NEXT(t) ) {
3226: NEXTND_pairs(s0,s);
3227: s->i1 = t->i1;
3228: s->i2 = t->i2;
1.14 noro 3229: SG(s) = SG(t);
1.61 noro 3230: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3231: }
1.37 noro 3232:
3233: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3234: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3235: old_red[i] = nd_red[i];
3236: nd_red[i] = 0;
3237: }
3238: for ( i = 0; i < REDTAB_LEN; i++ )
3239: for ( r = old_red[i]; r; r = NEXT(r) ) {
3240: NEWRHist(mr);
1.13 noro 3241: mr->index = r->index;
1.20 noro 3242: SG(mr) = SG(r);
1.61 noro 3243: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3244: h = ndl_hash_value(DL(mr));
3245: NEXT(mr) = nd_red[h];
3246: nd_red[h] = mr;
1.6 noro 3247: }
1.37 noro 3248: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3249: old_red = 0;
1.11 noro 3250: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3251: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3252: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3253: nd_psh[i] = r;
1.11 noro 3254: }
1.1 noro 3255: if ( s0 ) NEXT(s) = 0;
3256: prev_nm_free_list = 0;
3257: prev_ndp_free_list = 0;
1.71 noro 3258: #if 0
1.1 noro 3259: GC_gcollect();
1.71 noro 3260: #endif
1.1 noro 3261: return s0;
3262: }
3263:
1.61 noro 3264: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3265: {
1.57 noro 3266: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3267: struct order_pair *op;
1.1 noro 3268:
3269: n = nd_nvar;
1.61 noro 3270: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3271: omask0 = (1<<obpe)-1;
1.34 noro 3272: TD(r) = TD(d);
1.41 noro 3273: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3274: if ( nd_blockmask ) {
3275: l = nd_blockmask->n;
3276: op = nd_blockmask->order_pair;
3277: for ( i = 1; i < nd_exporigin; i++ )
3278: r[i] = d[i];
3279: for ( j = 0, s = 0; j < l; j++ ) {
3280: ord_l = op[j].length;
1.57 noro 3281: for ( i = 0; i < ord_l; i++, s++ ) {
3282: ei = GET_EXP_OLD(d,s);
3283: PUT_EXP(r,s,ei);
3284: }
1.1 noro 3285: }
1.43 noro 3286: } else {
1.56 noro 3287: for ( i = 0; i < n; i++ ) {
3288: ei = GET_EXP_OLD(d,i);
3289: PUT_EXP(r,i,ei);
3290: }
1.1 noro 3291: }
3292: }
1.3 noro 3293:
1.6 noro 3294: ND nd_copy(ND p)
3295: {
3296: NM m,mr,mr0;
1.41 noro 3297: int c,n;
1.6 noro 3298: ND r;
3299:
3300: if ( !p )
3301: return 0;
3302: else {
3303: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3304: NEXTNM(mr0,mr);
1.14 noro 3305: CM(mr) = CM(m);
3306: ndl_copy(DL(m),DL(mr));
1.6 noro 3307: }
3308: NEXT(mr) = 0;
1.31 noro 3309: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3310: SG(r) = SG(p);
1.6 noro 3311: return r;
3312: }
3313: }
3314:
1.53 noro 3315: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3316: {
1.148 ! noro 3317: NM m1,m2;
1.11 noro 3318: NDV p1,p2;
3319: ND t1,t2;
1.61 noro 3320: UINT *lcm;
1.148 ! noro 3321: P gp,tp;
! 3322: Q g,t;
1.31 noro 3323: int td;
1.11 noro 3324:
1.77 noro 3325: if ( !mod && nd_demand ) {
3326: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3327: } else {
1.77 noro 3328: if ( trace ) {
3329: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3330: } else {
3331: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3332: }
1.20 noro 3333: }
1.34 noro 3334: lcm = LCM(p);
1.148 ! noro 3335: NEWNM(m1); ndl_sub(lcm,HDL(p1),DL(m1));
! 3336: if ( ndl_check_bound2(p->i1,DL(m1)) ) {
! 3337: FREENM(m1); return 0;
! 3338: }
! 3339: NEWNM(m2); ndl_sub(lcm,HDL(p2),DL(m2));
! 3340: if ( ndl_check_bound2(p->i2,DL(m2)) ) {
! 3341: FREENM(m1); FREENM(m2); return 0;
! 3342: }
! 3343:
! 3344: if ( mod == -1 ) {
! 3345: CM(m1) = HCM(p2); CM(m2) = _chsgnsf(HCM(p1));
! 3346: } else if ( mod ) {
! 3347: CM(m1) = HCM(p2); CM(m2) = mod-HCM(p1);
! 3348: } else if ( nd_vc ) {
! 3349: ezgcdpz(nd_vc,HCP(p1),HCP(p2),&gp);
! 3350: divsp(nd_vc,HCP(p2),gp,&CP(m1));
! 3351: divsp(nd_vc,HCP(p1),gp,&tp); chsgnp(tp,&CP(m2));
! 3352: } else {
! 3353: igcd_cofactor(HCQ(p1),HCQ(p2),&g,&t,&CQ(m1)); chsgnq(t,&CQ(m2));
1.11 noro 3354: }
1.148 ! noro 3355: t1 = ndv_mul_nm(mod,m1,p1); t2 = ndv_mul_nm(mod,m2,p2);
1.31 noro 3356: *rp = nd_add(mod,t1,t2);
1.148 ! noro 3357: FREENM(m1); FREENM(m2);
1.11 noro 3358: return 1;
3359: }
3360:
1.19 noro 3361: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3362: {
3363: NMV m;
3364: int c,c1,len,i;
3365:
1.34 noro 3366: if ( !p ) return;
1.14 noro 3367: len = LEN(p);
1.71 noro 3368: if ( mod == -1 )
3369: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3370: CM(m) = _mulsf(CM(m),mul);
3371: else
3372: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3373: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3374: }
1.11 noro 3375: }
3376:
1.113 noro 3377: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3378: {
3379: NMV m;
1.113 noro 3380: Q c;
1.16 noro 3381: int len,i;
3382:
1.34 noro 3383: if ( !p ) return;
1.16 noro 3384: len = LEN(p);
3385: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3386: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3387: }
3388: }
3389:
1.55 noro 3390: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3391: int n2,i,j,l,n,tlen;
1.61 noro 3392: UINT *d0;
1.55 noro 3393: NM *tab,*psum;
3394: ND s,r;
3395: NM t;
3396: NMV m1;
3397:
3398: if ( !p ) return 0;
3399: n = NV(p); n2 = n>>1;
3400: d0 = DL(m0);
3401: l = LEN(p);
3402: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3403: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3404: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3405: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3406: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3407: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3408: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3409: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3410: for ( j = 0; j < tlen; j++ ) {
3411: if ( tab[j] ) {
3412: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3413: }
3414: }
3415: }
3416: for ( i = tlen-1, r = 0; i >= 0; i-- )
3417: if ( psum[i] ) {
3418: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3419: MKND(n,psum[i],j,s);
3420: r = nd_add(mod,r,s);
3421: }
1.56 noro 3422: if ( r ) SG(r) = SG(p)+TD(d0);
3423: return r;
1.55 noro 3424: }
3425:
1.56 noro 3426: /* product of monomials */
3427: /* XXX block order is not handled correctly */
3428:
1.55 noro 3429: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3430: {
1.56 noro 3431: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3432: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3433: Q *ctab_q;
3434: Q q,q1;
1.61 noro 3435: UINT c0,c1,c;
1.55 noro 3436: NM *p;
3437: NM m,t;
3438:
3439: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3440: if ( !m0 || !m1 ) return;
3441: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3442: NEWNM(m); d = DL(m);
1.56 noro 3443: if ( mod ) {
3444: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3445: } else
1.113 noro 3446: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3447: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3448: homo = n&1 ? 1 : 0;
3449: if ( homo ) {
3450: /* offset of h-degree */
3451: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3452: PUT_EXP(DL(m),n-1,h);
3453: TD(DL(m)) = h;
1.61 noro 3454: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3455: }
3456: tab[0] = m;
3457: NEWNM(m); d = DL(m);
1.57 noro 3458: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3459: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3460: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3461: /* xi^a*(Di^k*xi^l)*Di^b */
3462: a += l; b += k;
1.56 noro 3463: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3464: if ( !k || !l ) {
3465: for ( j = 0; j < curlen; j++ )
1.56 noro 3466: if ( t = tab[j] ) {
3467: dt = DL(t);
3468: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3469: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3470: }
3471: curlen *= k+1;
3472: continue;
3473: }
3474: min = MIN(k,l);
1.56 noro 3475: if ( mod ) {
1.61 noro 3476: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3477: mkwcm(k,l,mod,ctab);
3478: } else {
1.113 noro 3479: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3480: mkwc(k,l,ctab_q);
1.56 noro 3481: }
1.57 noro 3482: for ( j = min; j >= 0; j-- ) {
1.56 noro 3483: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3484: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3485: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3486: if ( homo ) {
3487: TD(d) = s;
1.56 noro 3488: PUT_EXP(d,n-1,s-h);
1.55 noro 3489: } else TD(d) = h;
1.61 noro 3490: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3491: if ( mod ) c = ctab[j];
1.113 noro 3492: else q = ctab_q[j];
1.57 noro 3493: p = tab+curlen*j;
3494: if ( j == 0 ) {
3495: for ( u = 0; u < curlen; u++, p++ ) {
3496: if ( tab[u] ) {
3497: ndl_addto(DL(tab[u]),d);
3498: if ( mod ) {
3499: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3500: } else {
1.113 noro 3501: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3502: }
3503: }
1.56 noro 3504: }
1.57 noro 3505: } else {
3506: for ( u = 0; u < curlen; u++, p++ ) {
3507: if ( tab[u] ) {
3508: NEWNM(t);
3509: ndl_add(DL(tab[u]),d,DL(t));
3510: if ( mod ) {
3511: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3512: } else
1.113 noro 3513: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3514: *p = t;
3515: }
1.55 noro 3516: }
3517: }
3518: }
3519: curlen *= k+1;
3520: }
3521: FREENM(m);
3522: }
3523:
1.63 noro 3524: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3525: {
3526: NM mr,mr0;
3527: NMV m;
3528: UINT *d,*dt,*dm;
3529: int c,n,td,i,c1,c2,len;
1.113 noro 3530: Q q;
1.63 noro 3531: ND r;
3532:
3533: if ( !p ) return 0;
3534: else {
3535: n = NV(p); m = BDY(p);
3536: d = DL(m0);
3537: len = LEN(p);
3538: mr0 = 0;
3539: td = TD(d);
3540: c = CM(m0);
3541: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3542: NEXTNM(mr0,mr);
3543: CM(mr) = 1;
3544: ndl_add(DL(m),d,DL(mr));
3545: }
3546: NEXT(mr) = 0;
3547: MKND(NV(p),mr0,len,r);
3548: SG(r) = SG(p) + TD(d);
3549: return r;
3550: }
3551: }
3552:
1.55 noro 3553: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3554: {
3555: NM mr,mr0;
3556: NMV m;
1.61 noro 3557: UINT *d,*dt,*dm;
1.9 noro 3558: int c,n,td,i,c1,c2,len;
1.146 noro 3559: P q;
1.9 noro 3560: ND r;
3561:
1.34 noro 3562: if ( !p ) return 0;
1.55 noro 3563: else if ( do_weyl )
1.71 noro 3564: if ( mod == -1 )
3565: error("ndv_mul_nm : not implemented (weyl)");
3566: else
3567: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3568: else {
3569: n = NV(p); m = BDY(p);
1.34 noro 3570: d = DL(m0);
1.14 noro 3571: len = LEN(p);
1.9 noro 3572: mr0 = 0;
1.34 noro 3573: td = TD(d);
1.71 noro 3574: if ( mod == -1 ) {
3575: c = CM(m0);
3576: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3577: NEXTNM(mr0,mr);
3578: CM(mr) = _mulsf(CM(m),c);
3579: ndl_add(DL(m),d,DL(mr));
3580: }
3581: } else if ( mod ) {
1.16 noro 3582: c = CM(m0);
3583: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3584: NEXTNM(mr0,mr);
3585: c1 = CM(m);
1.19 noro 3586: DMAR(c1,c,0,mod,c2);
1.16 noro 3587: CM(mr) = c2;
3588: ndl_add(DL(m),d,DL(mr));
3589: }
3590: } else {
1.146 noro 3591: q = CP(m0);
1.16 noro 3592: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3593: NEXTNM(mr0,mr);
1.146 noro 3594: mulp(nd_vc,CP(m),q,&CP(mr));
1.16 noro 3595: ndl_add(DL(m),d,DL(mr));
3596: }
1.4 noro 3597: }
1.9 noro 3598: NEXT(mr) = 0;
1.31 noro 3599: MKND(NV(p),mr0,len,r);
1.34 noro 3600: SG(r) = SG(p) + TD(d);
1.9 noro 3601: return r;
1.4 noro 3602: }
3603: }
3604:
1.104 noro 3605: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3606: {
3607: NM mq0,mq;
1.102 noro 3608: NMV tm;
1.113 noro 3609: Q q;
1.104 noro 3610: int i,nv,sg,c,c1,c2,hindex;
3611: ND p,t,r;
3612: N tnm;
3613:
1.124 noro 3614: if ( bucket->m < 0 ) return 0;
1.99 noro 3615: else {
1.104 noro 3616: nv = NV(d);
1.99 noro 3617: mq0 = 0;
1.102 noro 3618: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3619: while ( 1 ) {
1.113 noro 3620: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3621: if ( hindex < 0 ) break;
3622: p = bucket->body[hindex];
1.99 noro 3623: NEXTNM(mq0,mq);
3624: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3625: ndl_copy(DL(mq),DL(tm));
3626: if ( mod ) {
3627: c1 = invm(HCM(d),mod); c2 = HCM(p);
3628: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3629: CM(tm) = mod-c;
3630: } else {
1.113 noro 3631: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3632: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3633: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3634: }
3635: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3636: bucket->body[hindex] = nd_remove_head(p);
3637: t = nd_remove_head(t);
3638: add_pbucket(mod,bucket,t);
3639: }
3640: if ( !mq0 )
3641: r = 0;
3642: else {
3643: NEXT(mq) = 0;
3644: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3645: MKND(nv,mq0,i,r);
3646: /* XXX */
3647: SG(r) = HTD(r);
1.99 noro 3648: }
3649: return r;
3650: }
3651: }
3652:
1.43 noro 3653: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3654: {
1.13 noro 3655: NMV m,mr,mr0,t;
3656: int len,i,k;
1.11 noro 3657:
1.61 noro 3658: if ( !p ) return;
3659: m = BDY(p); len = LEN(p);
3660: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3661: m = (NMV)((char *)mr0+(len-1)*oadv);
3662: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3663: t = (NMV)ALLOCA(nmv_adv);
3664: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3665: CQ(t) = CQ(m);
1.61 noro 3666: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3667: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3668: CQ(mr) = CQ(t);
1.61 noro 3669: ndl_copy(DL(t),DL(mr));
3670: }
3671: BDY(p) = mr0;
3672: }
3673:
3674: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3675: {
3676: NMV m,mr,mr0;
3677: int len,i;
3678: NDV r;
1.11 noro 3679:
1.61 noro 3680: if ( !p ) return 0;
3681: m = BDY(p); len = LEN(p);
3682: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3683: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3684: ndl_zero(DL(mr));
3685: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3686: CQ(mr) = CQ(m);
1.11 noro 3687: }
1.61 noro 3688: MKNDV(NV(p),mr0,len,r);
3689: SG(r) = SG(p);
3690: return r;
1.11 noro 3691: }
3692:
1.61 noro 3693: /* duplicate p */
3694:
3695: NDV ndv_dup(int mod,NDV p)
1.3 noro 3696: {
3697: NDV d;
1.61 noro 3698: NMV t,m,m0;
1.3 noro 3699: int i,len;
3700:
1.34 noro 3701: if ( !p ) return 0;
1.31 noro 3702: len = LEN(p);
1.34 noro 3703: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3704: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3705: ndl_copy(DL(t),DL(m));
1.113 noro 3706: CQ(m) = CQ(t);
1.3 noro 3707: }
3708: MKNDV(NV(p),m0,len,d);
1.23 noro 3709: SG(d) = SG(p);
3710: return d;
3711: }
3712:
1.63 noro 3713: ND nd_dup(ND p)
3714: {
3715: ND d;
3716: NM t,m,m0;
3717:
3718: if ( !p ) return 0;
3719: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3720: NEXTNM(m0,m);
3721: ndl_copy(DL(t),DL(m));
1.113 noro 3722: CQ(m) = CQ(t);
1.63 noro 3723: }
3724: if ( m0 ) NEXT(m) = 0;
3725: MKND(NV(p),m0,LEN(p),d);
3726: SG(d) = SG(p);
3727: return d;
3728: }
3729:
1.61 noro 3730: /* XXX if p->len == 0 then it represents 0 */
3731:
3732: void ndv_mod(int mod,NDV p)
3733: {
3734: NMV t,d;
1.125 noro 3735: int r,s,u;
1.61 noro 3736: int i,len,dlen;
1.147 noro 3737: P cp;
1.146 noro 3738: Q c;
1.71 noro 3739: Obj gfs;
1.61 noro 3740:
3741: if ( !p ) return;
3742: len = LEN(p);
3743: dlen = 0;
1.71 noro 3744: if ( mod == -1 )
3745: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3746: simp_ff((Obj)CP(t),&gfs);
3747: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3748: CM(d) = r;
3749: ndl_copy(DL(t),DL(d));
3750: NMV_ADV(d);
3751: dlen++;
3752: }
1.71 noro 3753: else
3754: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.146 noro 3755: if ( nd_vc ) {
1.147 noro 3756: nd_subst_vector(nd_vc,CP(t),nd_subst,&cp);
1.146 noro 3757: c = (Q)cp;
3758: } else
3759: c = CQ(t);
3760: r = rem(NM(c),mod);
1.71 noro 3761: if ( r ) {
1.146 noro 3762: if ( SGN(c) < 0 )
1.113 noro 3763: r = mod-r;
1.146 noro 3764: if ( DN(c) ) {
3765: s = rem(DN(c),mod);
1.125 noro 3766: if ( !s )
3767: error("ndv_mod : division by 0");
3768: s = invm(s,mod);
3769: DMAR(r,s,0,mod,u); r = u;
3770: }
1.71 noro 3771: CM(d) = r;
3772: ndl_copy(DL(t),DL(d));
3773: NMV_ADV(d);
3774: dlen++;
3775: }
3776: }
1.61 noro 3777: LEN(p) = dlen;
3778: }
3779:
3780: NDV ptondv(VL vl,VL dvl,P p)
3781: {
3782: ND nd;
3783:
3784: nd = ptond(vl,dvl,p);
3785: return ndtondv(0,nd);
3786: }
3787:
3788: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3789: {
1.61 noro 3790: int n,i,j,k,e;
3791: VL tvl;
3792: V v;
3793: DCP dc;
3794: DCP *w;
3795: ND r,s,t,u;
3796: P x;
3797: int c;
3798: UINT *d;
1.23 noro 3799: NM m,m0;
1.61 noro 3800:
3801: if ( !p )
3802: return 0;
3803: else if ( NUM(p) ) {
3804: NEWNM(m);
3805: ndl_zero(DL(m));
1.113 noro 3806: CQ(m) = (Q)p;
1.61 noro 3807: NEXT(m) = 0;
3808: MKND(nd_nvar,m,1,r);
3809: SG(r) = 0;
3810: return r;
3811: } else {
3812: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3813: w = (DCP *)ALLOCA(k*sizeof(DCP));
3814: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3815: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3816: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3817: if ( !tvl ) {
3818: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3819: t = ptond(vl,dvl,COEF(w[j]));
3820: pwrp(vl,x,DEG(w[j]),&p);
3821: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3822: }
3823: return s;
3824: } else {
3825: NEWNM(m0); d = DL(m0);
3826: for ( j = k-1, s = 0; j >= 0; j-- ) {
3827: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3828: TD(d) = MUL_WEIGHT(e,i);
3829: if ( nd_blockmask) ndl_weight_mask(d);
3830: t = ptond(vl,dvl,COEF(w[j]));
3831: for ( m = BDY(t); m; m = NEXT(m) )
3832: ndl_addto(DL(m),d);
3833: SG(t) += TD(d);
3834: s = nd_add(0,s,t);
3835: }
3836: FREENM(m0);
3837: return s;
3838: }
3839: }
3840: }
3841:
3842: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3843: {
3844: VL tvl;
3845: int len,n,j,i,e;
3846: NMV m;
3847: Q q;
3848: P c;
3849: UINT *d;
3850: P s,r,u,t,w;
1.71 noro 3851: GFS gfs;
1.23 noro 3852:
1.34 noro 3853: if ( !p ) return 0;
1.61 noro 3854: else {
3855: len = LEN(p);
3856: n = NV(p);
3857: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3858: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3859: if ( mod == -1 ) {
3860: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3861: } else if ( mod ) {
1.61 noro 3862: STOQ(CM(m),q); c = (P)q;
3863: } else
1.113 noro 3864: c = CP(m);
1.61 noro 3865: d = DL(m);
3866: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3867: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3868: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3869: }
3870: addp(vl,s,t,&u); s = u;
3871: }
3872: return s;
1.23 noro 3873: }
1.3 noro 3874: }
3875:
1.61 noro 3876: NDV ndtondv(int mod,ND p)
1.11 noro 3877: {
3878: NDV d;
1.61 noro 3879: NMV m,m0;
3880: NM t;
3881: int i,len;
1.11 noro 3882:
1.34 noro 3883: if ( !p ) return 0;
1.61 noro 3884: len = LEN(p);
1.74 noro 3885: if ( mod )
3886: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3887: else
3888: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3889: #if 0
1.74 noro 3890: ndv_alloc += nmv_adv*len;
1.103 noro 3891: #endif
1.61 noro 3892: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3893: ndl_copy(DL(t),DL(m));
1.113 noro 3894: CQ(m) = CQ(t);
1.11 noro 3895: }
1.61 noro 3896: MKNDV(NV(p),m0,len,d);
1.14 noro 3897: SG(d) = SG(p);
1.11 noro 3898: return d;
3899: }
3900:
1.61 noro 3901: ND ndvtond(int mod,NDV p)
1.11 noro 3902: {
1.61 noro 3903: ND d;
3904: NM m,m0;
1.11 noro 3905: NMV t;
1.61 noro 3906: int i,len;
1.11 noro 3907:
1.34 noro 3908: if ( !p ) return 0;
1.11 noro 3909: m0 = 0;
1.61 noro 3910: len = p->len;
3911: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3912: NEXTNM(m0,m);
3913: ndl_copy(DL(t),DL(m));
1.113 noro 3914: CQ(m) = CQ(t);
1.11 noro 3915: }
3916: NEXT(m) = 0;
1.61 noro 3917: MKND(NV(p),m0,len,d);
1.14 noro 3918: SG(d) = SG(p);
1.11 noro 3919: return d;
3920: }
3921:
1.3 noro 3922: void ndv_print(NDV p)
3923: {
3924: NMV m;
3925: int i,len;
3926:
1.34 noro 3927: if ( !p ) printf("0\n");
1.3 noro 3928: else {
1.14 noro 3929: len = LEN(p);
1.3 noro 3930: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3931: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3932: else printf("+%d*",CM(m));
1.16 noro 3933: ndl_print(DL(m));
3934: }
3935: printf("\n");
3936: }
3937: }
3938:
1.113 noro 3939: void ndv_print_q(NDV p)
1.16 noro 3940: {
3941: NMV m;
3942: int i,len;
3943:
1.34 noro 3944: if ( !p ) printf("0\n");
1.16 noro 3945: else {
3946: len = LEN(p);
3947: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3948: printf("+");
1.113 noro 3949: printexpr(CO,(Obj)CQ(m));
1.16 noro 3950: printf("*");
1.14 noro 3951: ndl_print(DL(m));
1.3 noro 3952: }
3953: printf("\n");
3954: }
1.25 noro 3955: }
3956:
1.61 noro 3957: NODE ndv_reducebase(NODE x)
1.27 noro 3958: {
3959: int len,i,j;
3960: NDV *w;
3961: NODE t,t0;
3962:
3963: len = length(x);
3964: w = (NDV *)ALLOCA(len*sizeof(NDV));
3965: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3966: for ( i = 0; i < len; i++ ) {
3967: for ( j = 0; j < i; j++ ) {
3968: if ( w[i] && w[j] )
3969: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3970: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3971: }
3972: }
3973: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3974: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3975: }
3976: NEXT(t) = 0; x = t0;
3977: return x;
1.11 noro 3978: }
1.32 noro 3979:
1.43 noro 3980: /* XXX incomplete */
3981:
1.32 noro 3982: void nd_init_ord(struct order_spec *ord)
3983: {
1.43 noro 3984: switch ( ord->id ) {
1.32 noro 3985: case 0:
1.43 noro 3986: switch ( ord->ord.simple ) {
3987: case 0:
3988: nd_dcomp = 1;
3989: nd_isrlex = 1;
3990: break;
3991: case 1:
3992: nd_dcomp = 1;
3993: nd_isrlex = 0;
3994: break;
3995: case 2:
3996: nd_dcomp = 0;
3997: nd_isrlex = 0;
1.45 noro 3998: ndl_compare_function = ndl_lex_compare;
1.58 noro 3999: break;
4000: case 11:
4001: /* XXX */
4002: nd_dcomp = 0;
4003: nd_isrlex = 1;
4004: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 4005: break;
4006: default:
4007: error("nd_gr : unsupported order");
4008: }
1.32 noro 4009: break;
4010: case 1:
1.96 noro 4011: /* block order */
1.43 noro 4012: /* XXX */
4013: nd_dcomp = -1;
1.32 noro 4014: nd_isrlex = 0;
1.45 noro 4015: ndl_compare_function = ndl_block_compare;
1.34 noro 4016: break;
1.43 noro 4017: case 2:
1.96 noro 4018: /* matrix order */
4019: /* XXX */
4020: nd_dcomp = -1;
4021: nd_isrlex = 0;
4022: nd_matrix_len = ord->ord.matrix.row;
4023: nd_matrix = ord->ord.matrix.matrix;
4024: ndl_compare_function = ndl_matrix_compare;
4025: break;
4026: case 3:
1.97 noro 4027: /* composite order */
4028: nd_dcomp = -1;
4029: nd_isrlex = 0;
4030: nd_worb_len = ord->ord.composite.length;
4031: nd_worb = ord->ord.composite.w_or_b;
4032: ndl_compare_function = ndl_composite_compare;
1.32 noro 4033: break;
4034: }
1.41 noro 4035: nd_ord = ord;
1.32 noro 4036: }
4037:
1.43 noro 4038: BlockMask nd_create_blockmask(struct order_spec *ord)
4039: {
4040: int n,i,j,s,l;
1.61 noro 4041: UINT *t;
1.43 noro 4042: BlockMask bm;
4043:
1.96 noro 4044: /* we only create mask table for block order */
4045: if ( ord->id != 1 )
1.43 noro 4046: return 0;
4047: n = ord->ord.block.length;
4048: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
4049: bm->n = n;
4050: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 4051: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 4052: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 4053: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 4054: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
4055: l = bm->order_pair[i].length;
4056: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
4057: }
4058: return bm;
1.57 noro 4059: }
4060:
4061: EPOS nd_create_epos(struct order_spec *ord)
4062: {
4063: int i,j,l,s,ord_l,ord_o;
4064: EPOS epos;
4065: struct order_pair *op;
4066:
4067: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
4068: switch ( ord->id ) {
4069: case 0:
4070: if ( nd_isrlex ) {
4071: for ( i = 0; i < nd_nvar; i++ ) {
4072: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
4073: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
4074: }
4075: } else {
4076: for ( i = 0; i < nd_nvar; i++ ) {
4077: epos[i].i = nd_exporigin + i/nd_epw;
4078: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4079: }
4080: }
4081: break;
4082: case 1:
4083: /* block order */
4084: l = ord->ord.block.length;
4085: op = ord->ord.block.order_pair;
4086: for ( j = 0, s = 0; j < l; j++ ) {
4087: ord_o = op[j].order;
4088: ord_l = op[j].length;
4089: if ( !ord_o )
4090: for ( i = 0; i < ord_l; i++ ) {
4091: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
4092: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
4093: }
4094: else
4095: for ( i = 0; i < ord_l; i++ ) {
4096: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4097: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4098: }
4099: s += ord_l;
4100: }
4101: break;
4102: case 2:
1.96 noro 4103: /* matrix order */
4104: case 3:
4105: /* composite order */
4106: for ( i = 0; i < nd_nvar; i++ ) {
4107: epos[i].i = nd_exporigin + i/nd_epw;
4108: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4109: }
4110: break;
1.57 noro 4111: }
4112: return epos;
1.43 noro 4113: }
1.59 noro 4114:
4115: /* external interface */
4116:
4117: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4118: {
1.61 noro 4119: NODE t,in0,in;
1.59 noro 4120: ND nd,nf;
1.61 noro 4121: NDV ndv;
4122: VL vv,tv;
4123: int stat,nvar,max,e;
1.113 noro 4124: union oNDC dn;
1.146 noro 4125: Q cont;
4126: P pp;
1.59 noro 4127:
1.116 noro 4128: if ( !f ) {
4129: *rp = 0;
4130: return;
4131: }
1.59 noro 4132: pltovl(v,&vv);
1.74 noro 4133: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4134:
4135: /* get the degree bound */
4136: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4137: for ( tv = vv; tv; tv = NEXT(tv) ) {
4138: e = getdeg(tv->v,(P)BDY(t));
4139: max = MAX(e,max);
4140: }
4141: for ( tv = vv; tv; tv = NEXT(tv) ) {
4142: e = getdeg(tv->v,f);
4143: max = MAX(e,max);
4144: }
4145:
1.59 noro 4146: nd_init_ord(ord);
1.61 noro 4147: nd_setup_parameters(nvar,max);
4148:
4149: /* conversion to ndv */
4150: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4151: NEXTNODE(in0,in);
1.146 noro 4152: ptozp((P)BDY(t),1,&cont,&pp);
4153: BDY(in) = (pointer)ptondv(CO,vv,pp);
1.69 noro 4154: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4155: }
4156: NEXTNODE(in0,in);
4157: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4158: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4159: NEXT(in) = 0;
4160:
1.128 noro 4161: /* dont sort */
4162: ndv_setup(m,0,in0,1);
1.59 noro 4163: nd_psn--;
4164: nd_scale=2;
4165: while ( 1 ) {
4166: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.146 noro 4167: stat = nd_nf(m,0,nd,nd_ps,1,0,&nf);
1.59 noro 4168: if ( !stat ) {
4169: nd_psn++;
1.103 noro 4170: nd_reconstruct(0,0);
1.59 noro 4171: nd_psn--;
4172: } else
4173: break;
4174: }
1.61 noro 4175: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4176: }
4177:
4178: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4179: {
4180: NM m;
4181: UINT *t,*s;
4182: int i;
4183:
4184: for ( i = 0; i < n; i++ ) r[i] = 0;
4185: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4186: t = DL(m);
4187: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4188: r[i] = CM(m);
4189: }
4190: for ( i = 0; !r[i]; i++ );
4191: return i;
4192: }
4193:
1.113 noro 4194: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4195: {
1.107 noro 4196: NM m;
1.74 noro 4197: UINT *t,*s;
1.107 noro 4198: int i;
1.74 noro 4199:
4200: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4201: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4202: t = DL(m);
4203: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4204: r[i] = CQ(m);
1.74 noro 4205: }
4206: for ( i = 0; !r[i]; i++ );
4207: return i;
4208: }
4209:
1.129 noro 4210: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4211: {
4212: NM m;
4213: NMV mr;
4214: UINT *d,*t,*s;
4215: NDV p;
4216: int i,j,len;
4217: Q *r;
4218:
4219: m = pair->mul;
4220: d = DL(m);
4221: p = nd_ps[pair->index];
4222: len = LEN(p);
4223: r = (Q *)CALLOC(n,sizeof(Q));
4224: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4225: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4226: ndl_add(d,DL(mr),t);
4227: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4228: r[i] = CQ(mr);
4229: }
4230: return r;
4231: }
4232:
1.67 noro 4233: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4234: {
4235: NM m;
4236: NMV mr;
4237: UINT *d,*t,*s;
4238: NDV p;
1.67 noro 4239: unsigned char *ivc;
4240: unsigned short *ivs;
1.81 noro 4241: UINT *v,*ivi,*s0v;
1.67 noro 4242: int i,j,len,prev,diff,cdiff;
4243: IndArray r;
1.64 noro 4244:
4245: m = pair->mul;
4246: d = DL(m);
4247: p = nd_ps[pair->index];
4248: len = LEN(p);
4249: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4250: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4251: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4252: ndl_add(d,DL(mr),t);
4253: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4254: v[j] = i;
4255: }
1.81 noro 4256: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4257: r->head = v[0];
4258: diff = 0;
4259: for ( i = 1; i < len; i++ ) {
4260: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4261: }
4262: if ( diff < 256 ) {
4263: r->width = 1;
4264: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4265: r->index.c = ivc;
4266: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4267: } else if ( diff < 65536 ) {
4268: r->width = 2;
4269: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4270: r->index.s = ivs;
4271: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4272: } else {
4273: r->width = 4;
4274: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4275: r->index.i = ivi;
4276: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4277: }
1.67 noro 4278: return r;
1.64 noro 4279: }
4280:
1.135 noro 4281: int compress_array(Q *svect,Q *cvect,int n)
4282: {
4283: int i,j;
4284:
4285: for ( i = j = 0; i < n; i++ )
4286: if ( svect[i] ) cvect[j++] = svect[i];
4287: return j;
4288: }
4289:
4290: void expand_array(Q *svect,Q *cvect,int n)
4291: {
4292: int i,j;
4293:
4294: for ( i = j = 0; j < n; i++ )
4295: if ( svect[i] ) svect[i] = cvect[j++];
4296: }
4297:
1.133 noro 4298: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4299: {
1.135 noro 4300: int i,j,k,len,pos,prev,nz;
1.113 noro 4301: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4302: IndArray ivect;
4303: unsigned char *ivc;
4304: unsigned short *ivs;
4305: unsigned int *ivi;
4306: NDV redv;
4307: NMV mr;
4308: NODE rp;
4309: int maxrs;
1.133 noro 4310: double hmag;
1.135 noro 4311: Q *cvect;
1.107 noro 4312:
4313: maxrs = 0;
1.134 noro 4314: for ( i = 0; i < col && !svect[i]; i++ );
4315: if ( i == col ) return maxrs;
4316: hmag = p_mag((P)svect[i])*nd_scale;
1.135 noro 4317: cvect = (Q *)ALLOCA(col*sizeof(Q));
1.107 noro 4318: for ( i = 0; i < nred; i++ ) {
4319: ivect = imat[i];
4320: k = ivect->head;
4321: if ( svect[k] ) {
4322: maxrs = MAX(maxrs,rp0[i]->sugar);
1.133 noro 4323: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
1.107 noro 4324: len = LEN(redv); mr = BDY(redv);
1.113 noro 4325: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4326: chsgnq(cs,&mcs);
4327: if ( !UNIQ(cr) ) {
4328: for ( j = 0; j < col; j++ ) {
4329: mulq(svect[j],cr,&c1); svect[j] = c1;
4330: }
4331: }
1.107 noro 4332: svect[k] = 0; prev = k;
4333: switch ( ivect->width ) {
4334: case 1:
4335: ivc = ivect->index.c;
4336: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4337: pos = prev+ivc[j]; prev = pos;
1.113 noro 4338: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4339: }
4340: break;
4341: case 2:
4342: ivs = ivect->index.s;
4343: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4344: pos = prev+ivs[j]; prev = pos;
1.113 noro 4345: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4346: }
4347: break;
4348: case 4:
4349: ivi = ivect->index.i;
4350: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4351: pos = prev+ivi[j]; prev = pos;
1.113 noro 4352: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4353: }
4354: break;
4355: }
1.134 noro 4356: for ( j = k+1; j < col && !svect[j]; j++ );
4357: if ( j == col ) break;
4358: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
1.135 noro 4359: nz = compress_array(svect,cvect,col);
1.146 noro 4360: removecont_array((P *)cvect,nz);
1.135 noro 4361: expand_array(svect,cvect,nz);
1.134 noro 4362: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
4363: }
1.107 noro 4364: }
1.133 noro 4365: }
1.135 noro 4366: nz = compress_array(svect,cvect,col);
1.146 noro 4367: removecont_array((P *)cvect,nz);
1.135 noro 4368: expand_array(svect,cvect,nz);
1.133 noro 4369: if ( DP_Print ) {
4370: fprintf(asir_out,"-"); fflush(asir_out);
1.107 noro 4371: }
4372: return maxrs;
4373: }
4374:
1.76 noro 4375: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4376: {
1.67 noro 4377: int i,j,k,len,pos,prev;
1.66 noro 4378: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4379: IndArray ivect;
4380: unsigned char *ivc;
4381: unsigned short *ivs;
4382: unsigned int *ivi;
1.65 noro 4383: NDV redv;
1.67 noro 4384: NMV mr;
1.65 noro 4385: NODE rp;
1.76 noro 4386: int maxrs;
1.65 noro 4387:
1.76 noro 4388: maxrs = 0;
1.74 noro 4389: for ( i = 0; i < nred; i++ ) {
1.65 noro 4390: ivect = imat[i];
1.67 noro 4391: k = ivect->head; svect[k] %= m;
1.65 noro 4392: if ( c = svect[k] ) {
1.76 noro 4393: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4394: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4395: len = LEN(redv); mr = BDY(redv);
4396: svect[k] = 0; prev = k;
4397: switch ( ivect->width ) {
4398: case 1:
4399: ivc = ivect->index.c;
4400: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4401: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4402: prev = pos;
4403: DMA(c1,c,c2,up,lo);
4404: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4405: } else svect[pos] = lo;
4406: }
4407: break;
4408: case 2:
4409: ivs = ivect->index.s;
4410: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4411: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4412: prev = pos;
4413: DMA(c1,c,c2,up,lo);
4414: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4415: } else svect[pos] = lo;
4416: }
4417: break;
4418: case 4:
4419: ivi = ivect->index.i;
4420: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4421: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4422: prev = pos;
4423: DMA(c1,c,c2,up,lo);
4424: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4425: } else svect[pos] = lo;
4426: }
4427: break;
1.65 noro 4428: }
4429: }
4430: }
1.66 noro 4431: for ( i = 0; i < col; i++ )
4432: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4433: return maxrs;
1.65 noro 4434: }
4435:
1.76 noro 4436: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4437: {
4438: int i,j,k,len,pos,prev;
4439: UINT c,c1,c2,c3,up,lo,dmy;
4440: IndArray ivect;
4441: unsigned char *ivc;
4442: unsigned short *ivs;
4443: unsigned int *ivi;
4444: NDV redv;
4445: NMV mr;
4446: NODE rp;
1.76 noro 4447: int maxrs;
1.72 noro 4448:
1.76 noro 4449: maxrs = 0;
1.74 noro 4450: for ( i = 0; i < nred; i++ ) {
1.72 noro 4451: ivect = imat[i];
4452: k = ivect->head; svect[k] %= m;
4453: if ( c = svect[k] ) {
1.76 noro 4454: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4455: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4456: len = LEN(redv); mr = BDY(redv);
4457: svect[k] = 0; prev = k;
4458: switch ( ivect->width ) {
4459: case 1:
4460: ivc = ivect->index.c;
4461: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4462: pos = prev+ivc[j]; prev = pos;
4463: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4464: }
4465: break;
4466: case 2:
4467: ivs = ivect->index.s;
4468: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4469: pos = prev+ivs[j]; prev = pos;
4470: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4471: }
4472: break;
4473: case 4:
4474: ivi = ivect->index.i;
4475: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4476: pos = prev+ivi[j]; prev = pos;
4477: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4478: }
4479: break;
4480: }
4481: }
4482: }
1.76 noro 4483: return maxrs;
1.72 noro 4484: }
4485:
1.65 noro 4486: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4487: {
4488: int j,k,len;
4489: UINT *p;
4490: UINT c;
4491: NDV r;
4492: NMV mr0,mr;
4493:
4494: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4495: if ( !len ) return 0;
4496: else {
1.74 noro 4497: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4498: #if 0
1.74 noro 4499: ndv_alloc += nmv_adv*len;
1.103 noro 4500: #endif
1.65 noro 4501: mr = mr0;
4502: p = s0vect;
4503: for ( j = k = 0; j < col; j++, p += nd_wpd )
4504: if ( !rhead[j] ) {
4505: if ( c = vect[k++] ) {
4506: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4507: }
4508: }
4509: MKNDV(nd_nvar,mr0,len,r);
4510: return r;
4511: }
4512: }
4513:
1.129 noro 4514: /* for preprocessed vector */
4515:
1.113 noro 4516: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4517: {
4518: int j,k,len;
4519: UINT *p;
1.113 noro 4520: Q c;
1.107 noro 4521: NDV r;
4522: NMV mr0,mr;
4523:
4524: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4525: if ( !len ) return 0;
4526: else {
1.109 noro 4527: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4528: #if 0
4529: ndv_alloc += nmv_adv*len;
4530: #endif
4531: mr = mr0;
4532: p = s0vect;
4533: for ( j = k = 0; j < col; j++, p += nd_wpd )
4534: if ( !rhead[j] ) {
4535: if ( c = vect[k++] ) {
1.113 noro 4536: if ( DN(c) )
4537: error("afo");
4538: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4539: }
4540: }
4541: MKNDV(nd_nvar,mr0,len,r);
4542: return r;
4543: }
4544: }
4545:
1.129 noro 4546: /* for plain vector */
4547:
4548: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4549: {
4550: int j,k,len;
4551: UINT *p;
4552: Q c;
4553: NDV r;
4554: NMV mr0,mr;
4555:
4556: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4557: if ( !len ) return 0;
4558: else {
4559: mr0 = (NMV)GC_malloc(nmv_adv*len);
4560: #if 0
4561: ndv_alloc += nmv_adv*len;
4562: #endif
4563: mr = mr0;
4564: p = s0vect;
4565: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4566: if ( c = vect[k] ) {
4567: if ( DN(c) )
4568: error("afo");
4569: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4570: }
4571: MKNDV(nd_nvar,mr0,len,r);
4572: return r;
4573: }
4574: }
4575:
1.133 noro 4576: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 4577: {
4578: ND_pairs t;
4579: NODE sp0,sp;
4580: int stat;
4581: ND spol;
4582:
4583: for ( t = l; t; t = NEXT(t) ) {
1.133 noro 4584: stat = nd_sp(m,trace,t,&spol);
1.65 noro 4585: if ( !stat ) return 0;
4586: if ( spol ) {
4587: add_pbucket_symbolic(bucket,spol);
4588: }
4589: }
1.68 noro 4590: return 1;
1.65 noro 4591: }
4592:
1.133 noro 4593: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 4594: {
4595: NODE rp0,rp;
4596: NM mul,head,s0,s;
1.76 noro 4597: int index,col,i,sugar;
1.65 noro 4598: RHist h;
4599: UINT *s0v,*p;
4600: NM_ind_pair pair;
4601: ND red;
1.133 noro 4602: NDV *ps;
1.65 noro 4603:
4604: s0 = 0; rp0 = 0; col = 0;
1.133 noro 4605: ps = trace?nd_ps_trace:nd_ps;
1.65 noro 4606: while ( 1 ) {
4607: head = remove_head_pbucket_symbolic(bucket);
4608: if ( !head ) break;
4609: if ( !s0 ) s0 = head;
4610: else NEXT(s) = head;
4611: s = head;
4612: index = ndl_find_reducer(DL(head));
4613: if ( index >= 0 ) {
4614: h = nd_psh[index];
4615: NEWNM(mul);
4616: ndl_sub(DL(head),DL(h),DL(mul));
4617: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.133 noro 4618: sugar = TD(DL(mul))+SG(ps[index]);
1.76 noro 4619: MKNM_ind_pair(pair,mul,index,sugar);
1.133 noro 4620: red = ndv_mul_nm_symbolic(mul,ps[index]);
1.65 noro 4621: add_pbucket_symbolic(bucket,nd_remove_head(red));
4622: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4623: }
4624: col++;
4625: }
1.72 noro 4626: if ( rp0 ) NEXT(rp) = 0;
4627: NEXT(s) = 0;
1.65 noro 4628: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4629: for ( i = 0, p = s0v, s = s0; i < col;
4630: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4631: *s0vect = s0v;
4632: *r = rp0;
4633: return col;
4634: }
4635:
1.69 noro 4636: NODE nd_f4(int m)
4637: {
4638: int i,nh,stat,index;
4639: NODE r,g;
4640: ND_pairs d,l,t;
4641: ND spol,red;
4642: NDV nf,redv;
4643: NM s0,s;
1.84 noro 4644: NODE rp0,srp0,nflist;
1.69 noro 4645: int nsp,nred,col,rank,len,k,j,a;
4646: UINT c;
1.74 noro 4647: UINT **spmat;
1.69 noro 4648: UINT *s0vect,*svect,*p,*v;
4649: int *colstat;
4650: IndArray *imat;
4651: int *rhead;
4652: int spcol,sprow;
4653: int sugar;
4654: PGeoBucket bucket;
4655: struct oEGT eg0,eg1,eg_f4;
4656:
1.103 noro 4657: #if 0
1.74 noro 4658: ndv_alloc = 0;
1.103 noro 4659: #endif
1.69 noro 4660: g = 0; d = 0;
4661: for ( i = 0; i < nd_psn; i++ ) {
4662: d = update_pairs(d,g,i);
4663: g = update_base(g,i);
4664: }
4665: while ( d ) {
4666: get_eg(&eg0);
4667: l = nd_minsugarp(d,&d);
4668: sugar = SG(l);
4669: bucket = create_pbucket();
1.133 noro 4670: stat = nd_sp_f4(m,0,l,bucket);
1.69 noro 4671: if ( !stat ) {
4672: for ( t = l; NEXT(t); t = NEXT(t) );
4673: NEXT(t) = d; d = l;
1.103 noro 4674: d = nd_reconstruct(0,d);
1.69 noro 4675: continue;
4676: }
1.86 noro 4677: if ( bucket->m < 0 ) continue;
1.133 noro 4678: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
1.69 noro 4679: if ( !col ) {
4680: for ( t = l; NEXT(t); t = NEXT(t) );
4681: NEXT(t) = d; d = l;
1.103 noro 4682: d = nd_reconstruct(0,d);
1.69 noro 4683: continue;
4684: }
4685: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4686: if ( DP_Print )
4687: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4688: sugar,eg_f4.exectime+eg_f4.gctime);
4689: if ( 1 )
1.133 noro 4690: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
1.74 noro 4691: else
1.133 noro 4692: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
1.69 noro 4693: /* adding new bases */
1.74 noro 4694: for ( r = nflist; r; r = NEXT(r) ) {
4695: nf = (NDV)BDY(r);
1.69 noro 4696: ndv_removecont(m,nf);
1.129 noro 4697: if ( !m && nd_nalg ) {
4698: ND nf1;
4699:
4700: nf1 = ndvtond(m,nf);
4701: nd_monic(0,&nf1);
4702: nd_removecont(m,nf1);
4703: nf = ndtondv(m,nf1);
4704: }
1.77 noro 4705: nh = ndv_newps(m,nf,0);
1.69 noro 4706: d = update_pairs(d,g,nh);
4707: g = update_base(g,nh);
4708: }
4709: }
4710: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4711: #if 0
1.74 noro 4712: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4713: #endif
1.69 noro 4714: return g;
4715: }
1.74 noro 4716:
1.133 noro 4717: NODE nd_f4_trace(int m)
4718: {
4719: int i,nh,stat,index;
4720: NODE r,g;
4721: ND_pairs d,l,l0,t;
4722: ND spol,red;
4723: NDV nf,redv,nfqv,nfv;
4724: NM s0,s;
4725: NODE rp0,srp0,nflist;
4726: int nsp,nred,col,rank,len,k,j,a;
4727: UINT c;
4728: UINT **spmat;
4729: UINT *s0vect,*svect,*p,*v;
4730: int *colstat;
4731: IndArray *imat;
4732: int *rhead;
4733: int spcol,sprow;
4734: int sugar;
4735: PGeoBucket bucket;
4736: struct oEGT eg0,eg1,eg_f4;
4737:
4738: g = 0; d = 0;
4739: for ( i = 0; i < nd_psn; i++ ) {
4740: d = update_pairs(d,g,i);
4741: g = update_base(g,i);
4742: }
4743: while ( d ) {
4744: get_eg(&eg0);
4745: l = nd_minsugarp(d,&d);
4746: sugar = SG(l);
4747: bucket = create_pbucket();
4748: stat = nd_sp_f4(m,0,l,bucket);
4749: if ( !stat ) {
4750: for ( t = l; NEXT(t); t = NEXT(t) );
4751: NEXT(t) = d; d = l;
4752: d = nd_reconstruct(1,d);
4753: continue;
4754: }
4755: if ( bucket->m < 0 ) continue;
4756: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
4757: if ( !col ) {
4758: for ( t = l; NEXT(t); t = NEXT(t) );
4759: NEXT(t) = d; d = l;
4760: d = nd_reconstruct(1,d);
4761: continue;
4762: }
4763: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4764: if ( DP_Print )
4765: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4766: sugar,eg_f4.exectime+eg_f4.gctime);
4767: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
4768: if ( !l0 ) continue;
4769: l = l0;
4770:
4771: /* over Q */
4772: bucket = create_pbucket();
4773: stat = nd_sp_f4(0,1,l,bucket);
4774: if ( !stat ) {
4775: for ( t = l; NEXT(t); t = NEXT(t) );
4776: NEXT(t) = d; d = l;
4777: d = nd_reconstruct(1,d);
4778: continue;
4779: }
4780: if ( bucket->m < 0 ) continue;
4781: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
4782: if ( !col ) {
4783: for ( t = l; NEXT(t); t = NEXT(t) );
4784: NEXT(t) = d; d = l;
4785: d = nd_reconstruct(1,d);
4786: continue;
4787: }
4788: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
4789: /* adding new bases */
4790: for ( r = nflist; r; r = NEXT(r) ) {
4791: nfqv = (NDV)BDY(r);
4792: ndv_removecont(0,nfqv);
4793: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
4794: if ( nd_nalg ) {
4795: ND nf1;
4796:
4797: nf1 = ndvtond(m,nfqv);
4798: nd_monic(0,&nf1);
4799: nd_removecont(0,nf1);
4800: nfqv = ndtondv(0,nf1); nd_free(nf1);
4801: }
4802: nfv = ndv_dup(0,nfqv);
4803: ndv_mod(m,nfv);
4804: ndv_removecont(m,nfv);
4805: nh = ndv_newps(0,nfv,nfqv);
4806: d = update_pairs(d,g,nh);
4807: g = update_base(g,nh);
4808: }
4809: }
4810: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
4811: #if 0
4812: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
4813: #endif
4814: return g;
4815: }
4816:
4817: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 4818: {
1.67 noro 4819: IndArray *imat;
1.106 noro 4820: int nsp,nred,i;
1.65 noro 4821: int *rhead;
1.106 noro 4822: NODE r0,rp;
1.74 noro 4823: ND_pairs sp;
4824: NM_ind_pair *rvect;
1.63 noro 4825:
1.74 noro 4826: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4827: nred = length(rp0);
1.74 noro 4828: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4829: rhead = (int *)ALLOCA(col*sizeof(int));
4830: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4831:
1.74 noro 4832: /* construction of index arrays */
4833: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4834: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4835: rvect[i] = (NM_ind_pair)BDY(rp);
4836: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4837: rhead[imat[i]->head] = 1;
4838: }
1.107 noro 4839: if ( m )
1.133 noro 4840: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
1.107 noro 4841: else
1.133 noro 4842: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4843: return r0;
4844: }
1.74 noro 4845:
1.106 noro 4846: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 4847: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 4848: {
4849: int spcol,sprow,a;
4850: int i,j,k,l,rank;
4851: NODE r0,r;
4852: ND_pairs sp;
4853: ND spol;
4854: int **spmat;
4855: UINT *svect,*v;
4856: int *colstat;
4857: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4858: int maxrs;
4859: int *spsugar;
1.133 noro 4860: ND_pairs *spactive;
1.106 noro 4861:
4862: spcol = col-nred;
4863: get_eg(&eg0);
1.74 noro 4864: /* elimination (1st step) */
4865: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4866: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4867: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.133 noro 4868: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
1.74 noro 4869: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4870: nd_sp(m,0,sp,&spol);
1.75 noro 4871: if ( !spol ) continue;
1.74 noro 4872: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4873: if ( m == -1 )
4874: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4875: else
4876: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4877: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4878: if ( i < col ) {
4879: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4880: for ( j = k = 0; j < col; j++ )
4881: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4882: spsugar[sprow] = MAX(maxrs,SG(spol));
1.133 noro 4883: if ( nz )
4884: spactive[sprow] = sp;
1.74 noro 4885: sprow++;
4886: }
1.76 noro 4887: nd_free(spol);
1.74 noro 4888: }
1.80 noro 4889: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4890: if ( DP_Print ) {
4891: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4892: fflush(asir_out);
4893: }
1.74 noro 4894: /* free index arrays */
4895: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4896:
4897: /* elimination (2nd step) */
4898: colstat = (int *)ALLOCA(spcol*sizeof(int));
4899: if ( m == -1 )
1.76 noro 4900: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4901: else
1.133 noro 4902: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
1.74 noro 4903: r0 = 0;
4904: for ( i = 0; i < rank; i++ ) {
4905: NEXTNODE(r0,r); BDY(r) =
4906: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4907: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4908: GC_free(spmat[i]);
4909: }
1.109 noro 4910: if ( r0 ) NEXT(r) = 0;
1.133 noro 4911:
1.74 noro 4912: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4913: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4914: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4915: if ( DP_Print ) {
1.80 noro 4916: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4917: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4918: nsp,nred,sprow,spcol,rank);
4919: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4920: }
1.133 noro 4921: if ( nz ) {
4922: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
4923: if ( rank > 0 ) {
4924: NEXT(spactive[rank-1]) = 0;
4925: *nz = spactive[0];
4926: } else
4927: *nz = 0;
4928: }
1.74 noro 4929: return r0;
4930: }
4931:
1.133 noro 4932: #if 1
4933: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 4934: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4935: {
4936: int spcol,sprow,a;
4937: int i,j,k,l,rank;
4938: NODE r0,r;
4939: ND_pairs sp;
4940: ND spol;
1.113 noro 4941: Q **spmat;
4942: Q *svect,*v;
1.107 noro 4943: int *colstat;
4944: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4945: int maxrs;
4946: int *spsugar;
1.137 noro 4947: pointer *w;
1.107 noro 4948:
4949: spcol = col-nred;
4950: get_eg(&eg0);
4951: /* elimination (1st step) */
1.113 noro 4952: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4953: svect = (Q *)ALLOCA(col*sizeof(Q));
4954: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4955: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
1.133 noro 4956: nd_sp(0,trace,sp,&spol);
1.107 noro 4957: if ( !spol ) continue;
1.113 noro 4958: nd_to_vect_q(s0vect,col,spol,svect);
1.133 noro 4959: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
1.107 noro 4960: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4961: if ( i < col ) {
1.113 noro 4962: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4963: for ( j = k = 0; j < col; j++ )
4964: if ( !rhead[j] ) v[k++] = svect[j];
4965: spsugar[sprow] = MAX(maxrs,SG(spol));
4966: sprow++;
4967: }
1.109 noro 4968: /* nd_free(spol); */
1.107 noro 4969: }
4970: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4971: if ( DP_Print ) {
4972: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4973: fflush(asir_out);
4974: }
4975: /* free index arrays */
1.109 noro 4976: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4977:
4978: /* elimination (2nd step) */
4979: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4980: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.137 noro 4981: w = (pointer *)ALLOCA(rank*sizeof(pointer));
4982: for ( i = 0; i < rank; i++ ) {
4983: w[rank-i-1] = (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
4984: SG((NDV)w[rank-i-1]) = spsugar[i];
4985: /* GC_free(spmat[i]); */
4986: }
1.138 noro 4987: #if 0
1.137 noro 4988: qsort(w,rank,sizeof(NDV),
4989: (int (*)(const void *,const void *))ndv_compare);
4990: #endif
1.107 noro 4991: r0 = 0;
4992: for ( i = 0; i < rank; i++ ) {
1.137 noro 4993: NEXTNODE(r0,r); BDY(r) = w[i];
1.107 noro 4994: }
1.137 noro 4995: if ( r0 ) NEXT(r) = 0;
1.135 noro 4996:
1.109 noro 4997: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 4998: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4999: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
5000: if ( DP_Print ) {
5001: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
5002: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
5003: nsp,nred,sprow,spcol,rank);
5004: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
5005: }
5006: return r0;
5007: }
1.129 noro 5008: #else
5009: void printm(Q **mat,int row,int col)
5010: {
5011: int i,j;
5012: printf("[");
5013: for ( i = 0; i < row; i++ ) {
5014: for ( j = 0; j < col; j++ ) {
5015: printexpr(CO,mat[i][j]); printf(" ");
5016: }
5017: printf("]\n");
5018: }
5019: }
5020:
5021: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
5022: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
5023: {
5024: int row,a;
5025: int i,j,rank;
5026: NODE r0,r;
5027: ND_pairs sp;
5028: ND spol;
5029: Q **mat;
5030: int *colstat;
5031: int *sugar;
5032:
5033: row = nsp+nred;
5034: /* make the matrix */
5035: mat = (Q **)ALLOCA(row*sizeof(Q *));
5036: sugar = (int *)ALLOCA(row*sizeof(int));
5037: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
5038: nd_sp(0,0,sp,&spol);
5039: if ( !spol ) continue;
5040: mat[row] = (Q *)MALLOC(col*sizeof(Q));
5041: nd_to_vect_q(s0vect,col,spol,mat[row]);
5042: sugar[row] = SG(spol);
5043: row++;
5044: }
5045: for ( i = 0; i < nred; i++, row++ ) {
5046: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
5047: sugar[row] = rvect[i]->sugar;
5048: }
5049: /* elimination */
5050: colstat = (int *)ALLOCA(col*sizeof(int));
5051: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
5052: r0 = 0;
5053: for ( i = 0; i < rank; i++ ) {
5054: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
5055: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
5056: if ( rhead[j] ) continue;
5057: NEXTNODE(r0,r); BDY(r) =
5058: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
5059: SG((NDV)BDY(r)) = sugar[i];
5060: }
5061: if ( r0 ) NEXT(r) = 0;
5062: printf("\n");
5063: return r0;
5064: }
5065: #endif
1.107 noro 5066:
1.74 noro 5067: FILE *nd_write,*nd_read;
5068:
5069: void nd_send_int(int a) {
5070: write_int(nd_write,&a);
5071: }
5072:
5073: void nd_send_intarray(int *p,int len) {
5074: write_intarray(nd_write,p,len);
5075: }
5076:
5077: int nd_recv_int() {
5078: int a;
5079:
5080: read_int(nd_read,&a);
5081: return a;
5082: }
5083:
5084: void nd_recv_intarray(int *p,int len) {
5085: read_intarray(nd_read,p,len);
5086: }
5087:
5088: void nd_send_ndv(NDV p) {
5089: int len,i;
5090: NMV m;
5091:
5092: if ( !p ) nd_send_int(0);
5093: else {
5094: len = LEN(p);
5095: nd_send_int(len);
5096: m = BDY(p);
5097: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5098: nd_send_int(CM(m));
5099: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 5100: }
1.74 noro 5101: }
5102: }
5103:
5104: void nd_send_nd(ND p) {
5105: int len,i;
5106: NM m;
5107:
5108: if ( !p ) nd_send_int(0);
5109: else {
5110: len = LEN(p);
5111: nd_send_int(len);
5112: m = BDY(p);
5113: for ( i = 0; i < len; i++, m = NEXT(m) ) {
5114: nd_send_int(CM(m));
5115: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 5116: }
1.74 noro 5117: }
5118: }
1.65 noro 5119:
1.74 noro 5120: NDV nd_recv_ndv()
5121: {
5122: int len,i;
5123: NMV m,m0;
5124: NDV r;
1.65 noro 5125:
1.74 noro 5126: len = nd_recv_int();
5127: if ( !len ) return 0;
5128: else {
5129: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5130: #if 0
1.74 noro 5131: ndv_alloc += len*nmv_adv;
1.103 noro 5132: #endif
1.74 noro 5133: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5134: CM(m) = nd_recv_int();
5135: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 5136: }
1.74 noro 5137: MKNDV(nd_nvar,m0,len,r);
5138: return r;
5139: }
5140: }
1.65 noro 5141:
1.74 noro 5142: int ox_exec_f4_red(Q proc)
5143: {
5144: Obj obj;
5145: STRING fname;
5146: NODE arg;
5147: int s;
5148: extern int ox_need_conv,ox_file_io;
5149:
5150: MKSTR(fname,"nd_exec_f4_red");
5151: arg = mknode(2,proc,fname);
5152: Pox_cmo_rpc(arg,&obj);
5153: s = get_ox_server_id(QTOS(proc));
5154: nd_write = iofp[s].out;
5155: nd_read = iofp[s].in;
5156: ox_need_conv = ox_file_io = 0;
5157: return s;
5158: }
5159:
1.133 noro 5160: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5161: {
5162: int nsp,nred;
5163: int i,rank,s;
5164: NODE rp,r0,r;
5165: ND_pairs sp;
5166: NM_ind_pair pair;
5167: NMV nmv;
5168: NM nm;
5169: NDV nf;
5170: Obj proc,dmy;
5171:
5172: ox_launch_main(0,0,&proc);
5173: s = ox_exec_f4_red((Q)proc);
5174:
5175: nd_send_int(m);
5176: nd_send_int(nd_nvar);
5177: nd_send_int(nd_bpe);
5178: nd_send_int(nd_wpd);
5179: nd_send_int(nmv_adv);
5180:
1.89 noro 5181: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 5182:
5183: nd_send_int(nd_psn);
5184: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5185:
5186: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5187: nd_send_int(nsp);
5188: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5189: nd_send_int(sp->i1); nd_send_int(sp->i2);
5190: }
5191:
5192: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5193:
5194: nred = length(rp0); nd_send_int(nred);
5195: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5196: pair = (NM_ind_pair)BDY(rp);
5197: nd_send_int(pair->index);
5198: nd_send_intarray(pair->mul->dl,nd_wpd);
5199: }
5200: fflush(nd_write);
5201: rank = nd_recv_int();
5202: fprintf(asir_out,"rank=%d\n",rank);
5203: r0 = 0;
5204: for ( i = 0; i < rank; i++ ) {
5205: nf = nd_recv_ndv();
5206: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5207: }
5208: Pox_shutdown(mknode(1,proc),&dmy);
5209: return r0;
5210: }
5211:
5212: /* server side */
5213:
5214: void nd_exec_f4_red_dist()
5215: {
5216: int m,i,nsp,col,s0size,nred,spcol,j,k;
5217: NM_ind_pair *rp0;
5218: NDV nf;
5219: UINT *s0vect;
5220: IndArray *imat;
5221: int *rhead;
5222: int **spmat;
5223: UINT *svect,*v;
5224: ND_pairs *sp0;
5225: int *colstat;
5226: int a,sprow,rank;
1.89 noro 5227: struct order_spec *ord;
1.74 noro 5228: Obj ordspec;
5229: ND spol;
1.76 noro 5230: int maxrs;
5231: int *spsugar;
1.74 noro 5232:
5233: nd_read = iofp[0].in;
5234: nd_write = iofp[0].out;
5235: m = nd_recv_int();
5236: nd_nvar = nd_recv_int();
5237: nd_bpe = nd_recv_int();
5238: nd_wpd = nd_recv_int();
5239: nmv_adv = nd_recv_int();
5240:
5241: loadobj(nd_read,&ordspec);
1.89 noro 5242: create_order_spec(0,ordspec,&ord);
5243: nd_init_ord(ord);
1.74 noro 5244: nd_setup_parameters(nd_nvar,0);
5245:
5246: nd_psn = nd_recv_int();
5247: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5248: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5249: for ( i = 0; i < nd_psn; i++ ) {
5250: nd_ps[i] = nd_recv_ndv();
5251: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5252: }
5253:
5254: nsp = nd_recv_int();
5255: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5256: for ( i = 0; i < nsp; i++ ) {
5257: NEWND_pairs(sp0[i]);
5258: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5259: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5260: }
5261:
5262: col = nd_recv_int();
5263: s0size = col*nd_wpd;
5264: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5265: nd_recv_intarray(s0vect,s0size);
5266:
5267: nred = nd_recv_int();
5268: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5269: for ( i = 0; i < nred; i++ ) {
5270: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5271: rp0[i]->index = nd_recv_int();
5272: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5273: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5274: }
5275:
5276: spcol = col-nred;
5277: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5278: rhead = (int *)MALLOC(col*sizeof(int));
5279: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5280:
5281: /* construction of index arrays */
5282: for ( i = 0; i < nred; i++ ) {
5283: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5284: rhead[imat[i]->head] = 1;
5285: }
5286:
5287: /* elimination (1st step) */
5288: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5289: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5290: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5291: for ( a = sprow = 0; a < nsp; a++ ) {
5292: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5293: if ( !spol ) continue;
1.74 noro 5294: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5295: if ( m == -1 )
5296: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5297: else
5298: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5299: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5300: if ( i < col ) {
5301: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5302: for ( j = k = 0; j < col; j++ )
5303: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5304: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5305: sprow++;
5306: }
1.76 noro 5307: nd_free(spol);
1.74 noro 5308: }
5309: /* elimination (2nd step) */
5310: colstat = (int *)ALLOCA(spcol*sizeof(int));
5311: if ( m == -1 )
1.76 noro 5312: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5313: else
1.133 noro 5314: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
1.74 noro 5315: nd_send_int(rank);
5316: for ( i = 0; i < rank; i++ ) {
5317: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5318: nd_send_ndv(nf);
1.63 noro 5319: }
1.74 noro 5320: fflush(nd_write);
1.107 noro 5321: }
5322:
1.113 noro 5323: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5324: {
1.109 noro 5325: int mod,i,j,t,c,rank,rank0,inv;
5326: int *ci,*ri;
1.113 noro 5327: Q dn;
1.109 noro 5328: MAT m,nm;
1.108 noro 5329: int **wmat;
5330:
5331: /* XXX */
5332: mod = 99999989;
5333: wmat = (int **)ALLOCA(row*sizeof(int *));
5334: for ( i = 0; i < row; i++ ) {
5335: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5336: for ( j = 0; j < col; j++ ) {
1.113 noro 5337: if ( mat0[i][j] ) {
5338: t = rem(NM(mat0[i][j]),mod);
5339: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5340: wmat[i][j] = t;
5341: } else
1.108 noro 5342: wmat[i][j] = 0;
5343: }
5344: }
1.133 noro 5345: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
1.109 noro 5346: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5347: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5348: if ( rank != rank0 )
5349: error("afo");
5350: for ( i = 0; i < row; i++ )
5351: for ( j = 0; j < col; j++ )
5352: mat0[i][j] = 0;
5353: c = col-rank;
5354: for ( i = 0; i < rank; i++ ) {
1.113 noro 5355: mat0[i][ri[i]] = dn;
1.109 noro 5356: for ( j = 0; j < c; j++ )
1.113 noro 5357: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5358: }
1.113 noro 5359: inv = invm(rem(NM(dn),mod),mod);
5360: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5361: for ( i = 0; i < row; i++ )
5362: for ( j = 0; j < col; j++ ) {
1.113 noro 5363: if ( mat0[i][j] ) {
5364: t = rem(NM(mat0[i][j]),mod);
5365: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5366: } else
1.109 noro 5367: t = 0;
5368: c = dmar(t,inv,0,mod);
5369: if ( wmat[i][j] != c )
5370: error("afo");
5371: }
5372: return rank;
1.76 noro 5373: }
5374:
1.133 noro 5375: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5376: {
5377: int i,j,k,l,inv,a,rank,s;
5378: unsigned int *t,*pivot,*pk;
5379: unsigned int **mat;
1.133 noro 5380: ND_pairs pair;
1.76 noro 5381:
5382: mat = (unsigned int **)mat0;
5383: for ( rank = 0, j = 0; j < col; j++ ) {
5384: for ( i = rank; i < row; i++ )
5385: mat[i][j] %= md;
5386: for ( i = rank; i < row; i++ )
5387: if ( mat[i][j] )
5388: break;
5389: if ( i == row ) {
5390: colstat[j] = 0;
5391: continue;
5392: } else
5393: colstat[j] = 1;
5394: if ( i != rank ) {
5395: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5396: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
1.133 noro 5397: if ( spactive ) {
5398: pair = spactive[i]; spactive[i] = spactive[rank];
5399: spactive[rank] = pair;
5400: }
1.76 noro 5401: }
5402: pivot = mat[rank];
5403: s = sugar[rank];
5404: inv = invm(pivot[j],md);
5405: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5406: if ( *pk ) {
5407: if ( *pk >= (unsigned int)md )
5408: *pk %= md;
5409: DMAR(*pk,inv,0,md,*pk)
5410: }
5411: for ( i = rank+1; i < row; i++ ) {
5412: t = mat[i];
5413: if ( a = t[j] ) {
5414: sugar[i] = MAX(sugar[i],s);
5415: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5416: }
5417: }
5418: rank++;
5419: }
5420: for ( j = col-1, l = rank-1; j >= 0; j-- )
5421: if ( colstat[j] ) {
5422: pivot = mat[l];
5423: s = sugar[l];
5424: for ( i = 0; i < l; i++ ) {
5425: t = mat[i];
5426: t[j] %= md;
5427: if ( a = t[j] ) {
5428: sugar[i] = MAX(sugar[i],s);
5429: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5430: }
5431: }
5432: l--;
5433: }
5434: for ( j = 0, l = 0; l < rank; j++ )
5435: if ( colstat[j] ) {
5436: t = mat[l];
5437: for ( k = j; k < col; k++ )
5438: if ( t[k] >= (unsigned int)md )
5439: t[k] %= md;
5440: l++;
5441: }
5442: return rank;
5443: }
5444:
5445: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5446: {
5447: int i,j,k,l,inv,a,rank,s;
5448: unsigned int *t,*pivot,*pk;
5449: unsigned int **mat;
5450:
5451: mat = (unsigned int **)mat0;
5452: for ( rank = 0, j = 0; j < col; j++ ) {
5453: for ( i = rank; i < row; i++ )
5454: if ( mat[i][j] )
5455: break;
5456: if ( i == row ) {
5457: colstat[j] = 0;
5458: continue;
5459: } else
5460: colstat[j] = 1;
5461: if ( i != rank ) {
5462: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5463: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5464: }
5465: pivot = mat[rank];
5466: s = sugar[rank];
5467: inv = _invsf(pivot[j]);
5468: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5469: if ( *pk )
5470: *pk = _mulsf(*pk,inv);
5471: for ( i = rank+1; i < row; i++ ) {
5472: t = mat[i];
5473: if ( a = t[j] ) {
5474: sugar[i] = MAX(sugar[i],s);
5475: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5476: }
5477: }
5478: rank++;
5479: }
5480: for ( j = col-1, l = rank-1; j >= 0; j-- )
5481: if ( colstat[j] ) {
5482: pivot = mat[l];
5483: s = sugar[l];
5484: for ( i = 0; i < l; i++ ) {
5485: t = mat[i];
5486: if ( a = t[j] ) {
5487: sugar[i] = MAX(sugar[i],s);
5488: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5489: }
5490: }
5491: l--;
5492: }
5493: return rank;
1.77 noro 5494: }
5495:
5496: int ndv_ishomo(NDV p)
5497: {
5498: NMV m;
5499: int len,h;
5500:
5501: if ( !p ) return 1;
5502: len = LEN(p);
5503: m = BDY(p);
5504: h = TD(DL(m));
5505: NMV_ADV(m);
5506: for ( len--; len; len--, NMV_ADV(m) )
5507: if ( TD(DL(m)) != h ) return 0;
5508: return 1;
5509: }
5510:
5511: void ndv_save(NDV p,int index)
5512: {
5513: FILE *s;
5514: char name[BUFSIZ];
5515: short id;
5516: int nv,sugar,len,n,i,td,e,j;
5517: NMV m;
5518: unsigned int *dl;
5519:
5520: sprintf(name,"%s/%d",Demand,index);
5521: s = fopen(name,"w");
5522: savevl(s,0);
5523: if ( !p ) {
5524: saveobj(s,0);
5525: return;
5526: }
5527: id = O_DP;
5528: nv = NV(p);
5529: sugar = SG(p);
5530: len = LEN(p);
5531: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5532: write_int(s,&len);
5533:
5534: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5535: saveobj(s,(Obj)CQ(m));
1.77 noro 5536: dl = DL(m);
5537: td = TD(dl);
5538: write_int(s,&td);
5539: for ( j = 0; j < nv; j++ ) {
5540: e = GET_EXP(dl,j);
5541: write_int(s,&e);
5542: }
5543: }
5544: fclose(s);
5545: }
5546:
5547: NDV ndv_load(int index)
5548: {
5549: FILE *s;
5550: char name[BUFSIZ];
5551: short id;
5552: int nv,sugar,len,n,i,td,e,j;
5553: NDV d;
5554: NMV m0,m;
5555: unsigned int *dl;
5556: Obj obj;
5557:
5558: sprintf(name,"%s/%d",Demand,index);
5559: s = fopen(name,"r");
5560: if ( !s ) return 0;
5561:
5562: skipvl(s);
5563: read_short(s,&id);
5564: if ( !id ) return 0;
5565: read_int(s,&nv);
5566: read_int(s,&sugar);
5567: read_int(s,&len);
5568:
5569: m0 = m = MALLOC(len*nmv_adv);
5570: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5571: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5572: dl = DL(m);
5573: ndl_zero(dl);
5574: read_int(s,&td); TD(dl) = td;
5575: for ( j = 0; j < nv; j++ ) {
5576: read_int(s,&e);
5577: PUT_EXP(dl,j,e);
5578: }
5579: if ( nd_blockmask ) ndl_weight_mask(dl);
5580: }
5581: fclose(s);
5582: MKNDV(nv,m0,len,d);
5583: SG(d) = sugar;
5584: return d;
1.99 noro 5585: }
5586:
1.102 noro 5587: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5588: {
5589: VL fv,tv;
1.102 noro 5590: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5591: pointer **m;
1.113 noro 5592: Q mone;
1.125 noro 5593: P **w;
1.126 noro 5594: P mp,r;
1.102 noro 5595: NDV **dm;
5596: NDV *t,*mi,*mj;
5597: NDV d,s,mij,mjj;
5598: ND u;
5599: NMV nmv;
1.114 noro 5600: UINT *bound;
1.99 noro 5601: PGeoBucket bucket;
5602: struct order_spec *ord;
1.126 noro 5603: Q dq,dt,ds;
5604: N gn,qn,dn0,nm,dn;
1.99 noro 5605:
5606: create_order_spec(0,0,&ord);
5607: nd_init_ord(ord);
5608: get_vars((Obj)f,&fv);
5609: if ( f->row != f->col )
5610: error("nd_det : non-square matrix");
5611: n = f->row;
1.125 noro 5612: m = f->body;
1.99 noro 5613: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5614:
5615: if ( !nvar ) {
5616: if ( !mod )
5617: detp(CO,(P **)m,n,rp);
5618: else {
5619: w = (P **)almat_pointer(n,n);
5620: for ( i = 0; i < n; i++ )
5621: for ( j = 0; j < n; j++ )
5622: ptomp(mod,(P)m[i][j],&w[i][j]);
5623: detmp(CO,mod,w,n,&mp);
5624: mptop(mp,rp);
5625: }
5626: return;
5627: }
1.126 noro 5628:
5629: if ( !mod ) {
5630: w = (P **)almat_pointer(n,n);
5631: dq = ONE;
5632: for ( i = 0; i < n; i++ ) {
5633: dn0 = ONEN;
5634: for ( j = 0; j < n; j++ ) {
5635: if ( !m[i][j] ) continue;
5636: lgp(m[i][j],&nm,&dn);
5637: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5638: }
5639: if ( !UNIN(dn0) ) {
5640: NTOQ(dn0,1,ds);
5641: for ( j = 0; j < n; j++ )
5642: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5643: mulq(dq,ds,&dt); dq = dt;
5644: } else
5645: for ( j = 0; j < n; j++ )
5646: w[i][j] = (P)m[i][j];
5647: }
5648: m = (pointer **)w;
5649: }
1.125 noro 5650:
1.99 noro 5651: for ( i = 0, max = 0; i < n; i++ )
5652: for ( j = 0; j < n; j++ )
5653: for ( tv = fv; tv; tv = NEXT(tv) ) {
5654: e = getdeg(tv->v,(P)m[i][j]);
5655: max = MAX(e,max);
5656: }
1.114 noro 5657: nd_setup_parameters(nvar,max);
1.102 noro 5658: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5659: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5660: for ( j = 0; j < n; j++ ) {
5661: dm[i][j] = ptondv(CO,fv,m[i][j]);
5662: if ( mod ) ndv_mod(mod,dm[i][j]);
5663: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5664: }
5665: d = ptondv(CO,fv,(P)ONE);
5666: if ( mod ) ndv_mod(mod,d);
1.113 noro 5667: chsgnq(ONE,&mone);
1.99 noro 5668: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5669: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5670: for ( i = j; i < n && !dm[i][j]; i++ );
5671: if ( i == n ) {
5672: *rp = 0;
5673: return;
5674: }
1.102 noro 5675: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5676: for ( k = j; k < n; k++ )
5677: for ( l = j; l < n; l++ )
1.102 noro 5678: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5679: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5680: }
5681: if ( k0 != j ) {
5682: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5683: sgn = -sgn;
5684: }
5685: if ( l0 != j ) {
5686: for ( k = j; k < n; k++ ) {
5687: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5688: }
5689: sgn = -sgn;
5690: }
1.114 noro 5691: bound = nd_det_compute_bound(dm,n,j);
5692: if ( ndl_check_bound(bound,bound) )
5693: nd_det_reconstruct(dm,n,j,d);
5694:
1.99 noro 5695: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5696: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5697: mi = dm[i]; mij = mi[j];
1.102 noro 5698: if ( mod )
5699: ndv_mul_c(mod,mij,mod-1);
5700: else
1.113 noro 5701: ndv_mul_c_q(mij,mone);
1.99 noro 5702: for ( k = j+1; k < n; k++ ) {
1.114 noro 5703: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5704: bucket = create_pbucket();
1.104 noro 5705: if ( mi[k] ) {
1.102 noro 5706: nmv = BDY(mjj); len = LEN(mjj);
5707: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5708: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5709: add_pbucket(mod,bucket,u);
1.99 noro 5710: }
1.104 noro 5711: }
1.99 noro 5712: if ( mj[k] && mij ) {
1.102 noro 5713: nmv = BDY(mij); len = LEN(mij);
5714: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5715: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5716: add_pbucket(mod,bucket,u);
1.99 noro 5717: }
5718: }
1.104 noro 5719: u = nd_quo(mod,bucket,d);
1.102 noro 5720: mi[k] = ndtondv(mod,u);
1.99 noro 5721: }
1.114 noro 5722: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5723: }
5724: d = mjj;
5725: }
1.125 noro 5726: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5727: if ( sgn < 0 )
1.102 noro 5728: if ( mod )
5729: ndv_mul_c(mod,d,mod-1);
5730: else
1.113 noro 5731: ndv_mul_c_q(d,mone);
1.126 noro 5732: r = ndvtop(mod,CO,fv,d);
5733: if ( !mod && !UNIQ(dq) )
5734: divsp(CO,r,(P)dq,rp);
5735: else
5736: *rp = r;
1.99 noro 5737: }
5738:
1.102 noro 5739: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5740: {
5741: NM mr,mr0;
1.102 noro 5742: NM tnm;
5743: NMV m;
1.99 noro 5744: UINT *d0,*dt,*dm;
5745: int c,n,td,i,c1,c2,len;
1.113 noro 5746: Q q;
1.99 noro 5747: ND r;
5748:
5749: if ( !p ) return 0;
5750: else {
1.102 noro 5751: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5752: d0 = DL(m0);
1.102 noro 5753: td = TD(d);
1.99 noro 5754: mr0 = 0;
5755: NEWNM(tnm);
1.102 noro 5756: if ( mod ) {
5757: c = CM(m0);
5758: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5759: ndl_add(DL(m),d0,DL(tnm));
5760: if ( ndl_reducible(DL(tnm),d) ) {
5761: NEXTNM(mr0,mr);
5762: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5763: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5764: }
5765: }
5766: } else {
1.113 noro 5767: q = CQ(m0);
1.102 noro 5768: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5769: ndl_add(DL(m),d0,DL(tnm));
5770: if ( ndl_reducible(DL(tnm),d) ) {
5771: NEXTNM(mr0,mr);
1.113 noro 5772: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5773: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5774: }
1.99 noro 5775: }
5776: }
5777: if ( !mr0 )
5778: return 0;
5779: else {
5780: NEXT(mr) = 0;
1.105 noro 5781: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5782: MKND(NV(p),mr0,len,r);
5783: SG(r) = SG(p) + TD(d0);
5784: return r;
5785: }
5786: }
1.114 noro 5787: }
5788:
5789: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5790: {
5791: int i,obpe,oadv,h,k,l;
5792: static NM prev_nm_free_list;
5793: EPOS oepos;
5794:
5795: obpe = nd_bpe;
5796: oadv = nmv_adv;
5797: oepos = nd_epos;
5798: if ( obpe < 2 ) nd_bpe = 2;
5799: else if ( obpe < 3 ) nd_bpe = 3;
5800: else if ( obpe < 4 ) nd_bpe = 4;
5801: else if ( obpe < 5 ) nd_bpe = 5;
5802: else if ( obpe < 6 ) nd_bpe = 6;
5803: else if ( obpe < 8 ) nd_bpe = 8;
5804: else if ( obpe < 10 ) nd_bpe = 10;
5805: else if ( obpe < 16 ) nd_bpe = 16;
5806: else if ( obpe < 32 ) nd_bpe = 32;
5807: else error("nd_det_reconstruct : exponent too large");
5808:
5809: nd_setup_parameters(nd_nvar,0);
5810: prev_nm_free_list = _nm_free_list;
5811: _nm_free_list = 0;
5812: for ( k = j; k < n; k++ )
5813: for (l = j; l < n; l++ )
5814: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5815: ndv_realloc(d,obpe,oadv,oepos);
5816: prev_nm_free_list = 0;
5817: #if 0
5818: GC_gcollect();
5819: #endif
5820: }
5821:
5822: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5823: {
5824: UINT *d0,*d1,*d,*t,*r;
5825: int k,l;
5826:
5827: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5828: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5829: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5830: for ( k = j; k < n; k++ )
5831: for ( l = j; l < n; l++ )
5832: if ( dm[k][l] ) {
5833: d = ndv_compute_bound(dm[k][l]);
5834: ndl_lcm(d,d0,d1);
5835: t = d1; d1 = d0; d0 = t;
5836: }
5837: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5838: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5839: return r;
1.117 noro 5840: }
5841:
5842: DL nd_separate_d(UINT *d,UINT *trans)
5843: {
1.121 noro 5844: int n,td,i,e,j;
1.117 noro 5845: DL a;
5846:
5847: ndl_zero(trans);
5848: td = 0;
1.121 noro 5849: for ( i = 0; i < nd_ntrans; i++ ) {
5850: e = GET_EXP(d,i);
5851: PUT_EXP(trans,i,e);
5852: td += MUL_WEIGHT(e,i);
5853: }
5854: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5855: /* homogenized */
5856: i = nd_nvar-1;
1.117 noro 5857: e = GET_EXP(d,i);
5858: PUT_EXP(trans,i,e);
5859: td += MUL_WEIGHT(e,i);
5860: }
5861: TD(trans) = td;
5862: if ( nd_blockmask) ndl_weight_mask(trans);
5863: NEWDL(a,nd_nalg);
5864: td = 0;
1.121 noro 5865: for ( i = 0; i < nd_nalg; i++ ) {
5866: j = nd_ntrans+i;
5867: e = GET_EXP(d,j);
5868: a->d[i] = e;
1.117 noro 5869: td += e;
5870: }
5871: a->td = td;
5872: return a;
5873: }
5874:
1.118 noro 5875: int nd_monic(int mod,ND *p)
1.117 noro 5876: {
5877: UINT *trans,*t;
5878: DL alg;
5879: MP mp0,mp;
1.118 noro 5880: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5881: ND r;
1.117 noro 5882: DL dl;
5883: DP nm;
5884: NDV ndv;
1.118 noro 5885: DAlg inv,cd;
1.117 noro 5886: ND s,c;
1.118 noro 5887: Q l,mul;
5888: N ln;
5889: int n,ntrans,i,e,td,is_lc,len;
5890: NumberField nf;
5891: struct oEGT eg0,eg1;
5892:
5893: if ( !(nf = get_numberfield()) )
5894: error("nd_monic : current_numberfield is not set");
1.117 noro 5895:
1.118 noro 5896: /* Q coef -> DAlg coef */
5897: NEWNM(ma0); ma = ma0;
5898: m = BDY(*p);
5899: is_lc = 1;
5900: while ( 1 ) {
5901: NEWMP(mp0); mp = mp0;
5902: mp->c = (P)CQ(m);
5903: mp->dl = nd_separate_d(DL(m),DL(ma));
5904: NEWNM(mb);
5905: for ( m = NEXT(m); m; m = NEXT(m) ) {
5906: alg = nd_separate_d(DL(m),DL(mb));
5907: if ( !ndl_equal(DL(ma),DL(mb)) )
5908: break;
5909: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5910: }
5911: NEXT(mp) = 0;
5912: MKDP(nd_nalg,mp0,nm);
5913: MKDAlg(nm,ONE,cd);
5914: if ( is_lc == 1 ) {
5915: /* if the lc is a rational number, we have nothing to do */
5916: if ( !mp0->dl->td )
5917: return 1;
5918:
5919: get_eg(&eg0);
5920: invdalg(cd,&inv);
5921: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5922: /* check the validity of inv */
5923: if ( mod && !rem(NM(inv->dn),mod) )
5924: return 0;
5925: CA(ma) = nf->one;
5926: is_lc = 0;
5927: ln = ONEN;
5928: } else {
5929: muldalg(cd,inv,&CA(ma));
5930: lcmn(ln,NM(CA(ma)->dn),&ln);
5931: }
5932: if ( m ) {
5933: NEXT(ma) = mb; ma = mb;
5934: } else {
5935: NEXT(ma) = 0;
5936: break;
5937: }
5938: }
5939: /* l = lcm(denoms) */
5940: NTOQ(ln,1,l);
5941: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5942: divq(l,CA(m)->dn,&mul);
5943: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5944: NEXTNM(mr0,mr);
5945: mulq((Q)mp->c,mul,&CQ(mr));
5946: dl = mp->dl;
5947: td = TD(DL(m));
5948: ndl_copy(DL(m),DL(mr));
1.121 noro 5949: for ( i = 0; i < nd_nalg; i++ ) {
5950: e = dl->d[i];
5951: PUT_EXP(DL(mr),i+nd_ntrans,e);
5952: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5953: }
5954: TD(DL(mr)) = td;
5955: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5956: }
1.117 noro 5957: }
1.118 noro 5958: NEXT(mr) = 0;
5959: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5960: MKND(NV(*p),mr0,len,r);
5961: /* XXX */
5962: SG(r) = SG(*p);
5963: nd_free(*p);
5964: *p = r;
5965: return 1;
1.59 noro 5966: }