Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.149
1.149 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.148 2006/11/28 02:17:54 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:
1.149 ! noro 55: extern int Denominator,DP_Multiple;
! 56:
1.1 noro 57: void nd_free_private_storage()
58: {
59: _nm_free_list = 0;
1.5 noro 60: _ndp_free_list = 0;
1.71 noro 61: #if 0
1.1 noro 62: GC_gcollect();
1.71 noro 63: #endif
1.1 noro 64: }
65:
66: void _NM_alloc()
67: {
68: NM p;
69: int i;
70:
1.11 noro 71: for ( i = 0; i < 1024; i++ ) {
1.61 noro 72: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 73: p->next = _nm_free_list; _nm_free_list = p;
74: }
75: }
76:
77: void _ND_alloc()
78: {
79: ND p;
80: int i;
81:
82: for ( i = 0; i < 1024; i++ ) {
83: p = (ND)GC_malloc(sizeof(struct oND));
84: p->body = (NM)_nd_free_list; _nd_free_list = p;
85: }
86: }
87:
88: void _NDP_alloc()
89: {
90: ND_pairs p;
91: int i;
92:
1.11 noro 93: for ( i = 0; i < 1024; i++ ) {
1.1 noro 94: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 95: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 96: p->next = _ndp_free_list; _ndp_free_list = p;
97: }
98: }
99:
1.30 noro 100: INLINE int nd_length(ND p)
1.1 noro 101: {
102: NM m;
103: int i;
104:
105: if ( !p )
106: return 0;
107: else {
108: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
109: return i;
110: }
111: }
112:
1.61 noro 113: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 114: {
1.61 noro 115: UINT u1,u2;
1.1 noro 116: int i,j;
117:
1.34 noro 118: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 119: #if USE_UNROLL
1.1 noro 120: switch ( nd_bpe ) {
1.62 noro 121: case 3:
122: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
123: u1 = d1[i]; u2 = d2[i];
124: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
125: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
126: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
127: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
128: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
129: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
130: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
131: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
132: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
133: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
134: }
135: return 1;
136: break;
1.1 noro 137: case 4:
1.41 noro 138: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 139: u1 = d1[i]; u2 = d2[i];
140: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 141: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
142: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
143: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
144: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
145: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
146: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
147: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 148: }
149: return 1;
150: break;
151: case 6:
1.41 noro 152: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 153: u1 = d1[i]; u2 = d2[i];
154: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 155: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
156: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
157: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
158: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 159: }
160: return 1;
161: break;
162: case 8:
1.41 noro 163: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 164: u1 = d1[i]; u2 = d2[i];
165: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 166: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
167: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
168: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 169: }
170: return 1;
171: break;
172: case 16:
1.41 noro 173: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 174: u1 = d1[i]; u2 = d2[i];
175: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 176: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 177: }
178: return 1;
179: break;
180: case 32:
1.41 noro 181: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 182: if ( d1[i] < d2[i] ) return 0;
183: return 1;
184: break;
185: default:
1.41 noro 186: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 187: u1 = d1[i]; u2 = d2[i];
188: for ( j = 0; j < nd_epw; j++ )
189: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
190: }
191: return 1;
192: }
1.65 noro 193: #else
194: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
195: u1 = d1[i]; u2 = d2[i];
196: for ( j = 0; j < nd_epw; j++ )
197: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
198: }
199: return 1;
200: #endif
1.1 noro 201: }
202:
1.61 noro 203: /*
204: * If the current order is a block order,
205: * then the last block is length 1 and contains
206: * the homo variable. Otherwise, the original
207: * order is either 0 or 2.
208: */
209:
210: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 211: {
1.61 noro 212: int w,i,e,n,omask0;
213:
214: omask0 = (1<<obpe)-1;
215: n = nd_nvar-1;
216: ndl_zero(r);
217: for ( i = 0; i < n; i++ ) {
218: e = GET_EXP_OLD(d,i);
219: PUT_EXP(r,i,e);
220: }
221: w = TD(d);
222: PUT_EXP(r,nd_nvar-1,weight-w);
223: TD(r) = weight;
224: if ( nd_blockmask ) ndl_weight_mask(r);
225: }
226:
227: void ndl_dehomogenize(UINT *d)
228: {
229: UINT mask;
230: UINT h;
1.31 noro 231: int i,bits;
1.23 noro 232:
1.44 noro 233: if ( nd_blockmask ) {
234: h = GET_EXP(d,nd_nvar-1);
1.45 noro 235: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 236: TD(d) -= h;
237: d[nd_exporigin-1] -= h;
238: } else {
239: if ( nd_isrlex ) {
240: if ( nd_bpe == 32 ) {
241: h = d[nd_exporigin];
242: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
243: d[i-1] = d[i];
244: d[i-1] = 0;
245: TD(d) -= h;
246: } else {
247: bits = nd_epw*nd_bpe;
248: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
249: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
250: for ( i = nd_exporigin; i < nd_wpd; i++ )
251: d[i] = ((d[i]<<nd_bpe)&mask)
252: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
253: TD(d) -= h;
254: }
1.45 noro 255: } else {
256: h = GET_EXP(d,nd_nvar-1);
257: XOR_EXP(d,nd_nvar-1,h);
258: TD(d) -= h;
259: }
1.44 noro 260: }
1.23 noro 261: }
262:
1.61 noro 263: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 264: {
1.61 noro 265: UINT t1,t2,u,u1,u2;
1.43 noro 266: int i,j,l;
1.1 noro 267:
1.65 noro 268: #if USE_UNROLL
1.1 noro 269: switch ( nd_bpe ) {
1.62 noro 270: case 3:
271: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
272: u1 = d1[i]; u2 = d2[i];
273: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
274: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
275: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
276: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
277: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
278: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
279: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
280: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
281: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
282: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
283: d[i] = u;
284: }
285: break;
1.1 noro 286: case 4:
1.41 noro 287: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 288: u1 = d1[i]; u2 = d2[i];
289: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 290: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
291: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
292: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
293: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
294: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
295: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
296: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 297: d[i] = u;
298: }
299: break;
300: case 6:
1.41 noro 301: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 302: u1 = d1[i]; u2 = d2[i];
303: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 304: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
305: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
306: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
307: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 308: d[i] = u;
309: }
310: break;
311: case 8:
1.41 noro 312: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 313: u1 = d1[i]; u2 = d2[i];
314: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 315: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
316: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
317: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 318: d[i] = u;
319: }
320: break;
321: case 16:
1.41 noro 322: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 323: u1 = d1[i]; u2 = d2[i];
324: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 325: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 326: d[i] = u;
327: }
328: break;
329: case 32:
1.41 noro 330: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 331: u1 = d1[i]; u2 = d2[i];
332: d[i] = u1>u2?u1:u2;
333: }
334: break;
335: default:
1.41 noro 336: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 337: u1 = d1[i]; u2 = d2[i];
338: for ( j = 0, u = 0; j < nd_epw; j++ ) {
339: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
340: }
341: d[i] = u;
342: }
343: break;
344: }
1.65 noro 345: #else
346: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
347: u1 = d1[i]; u2 = d2[i];
348: for ( j = 0, u = 0; j < nd_epw; j++ ) {
349: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
350: }
351: d[i] = u;
352: }
353: #endif
1.39 noro 354: TD(d) = ndl_weight(d);
1.61 noro 355: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 356: }
357:
1.61 noro 358: int ndl_weight(UINT *d)
1.1 noro 359: {
1.61 noro 360: UINT t,u;
1.1 noro 361: int i,j;
362:
1.60 noro 363: if ( current_dl_weight_vector )
364: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
365: u = GET_EXP(d,i);
366: t += MUL_WEIGHT(u,i);
367: }
368: else
369: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
370: u = d[i];
371: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
372: t += (u&nd_mask0);
373: }
1.1 noro 374: return t;
375: }
376:
1.61 noro 377: void ndl_weight_mask(UINT *d)
1.43 noro 378: {
1.61 noro 379: UINT t,u;
380: UINT *mask;
381: int i,j,k,l;
1.43 noro 382:
1.61 noro 383: l = nd_blockmask->n;
384: for ( k = 0; k < l; k++ ) {
385: mask = nd_blockmask->mask[k];
386: if ( current_dl_weight_vector )
387: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
388: u = GET_EXP_MASK(d,i,mask);
389: t += MUL_WEIGHT(u,i);
390: }
391: else
392: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
393: u = d[i]&mask[i];
394: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
395: t += (u&nd_mask0);
396: }
397: d[k+1] = t;
398: }
1.43 noro 399: }
400:
1.61 noro 401: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 402: {
403: int i;
404:
1.41 noro 405: d1 += nd_exporigin;
406: d2 += nd_exporigin;
407: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 408: if ( *d1 > *d2 )
1.32 noro 409: return nd_isrlex ? -1 : 1;
1.1 noro 410: else if ( *d1 < *d2 )
1.32 noro 411: return nd_isrlex ? 1 : -1;
1.1 noro 412: return 0;
413: }
414:
1.61 noro 415: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 416: {
417: int i,l,j,ord_o,ord_l;
418: struct order_pair *op;
1.61 noro 419: UINT t1,t2,m;
420: UINT *mask;
1.43 noro 421:
422: l = nd_blockmask->n;
423: op = nd_blockmask->order_pair;
424: for ( j = 0; j < l; j++ ) {
425: mask = nd_blockmask->mask[j];
426: ord_o = op[j].order;
427: if ( ord_o < 2 )
1.44 noro 428: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
429: else if ( t1 < t2 ) return -1;
1.43 noro 430: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 431: m = mask[i];
432: t1 = d1[i]&m;
433: t2 = d2[i]&m;
1.43 noro 434: if ( t1 > t2 )
435: return !ord_o ? -1 : 1;
436: else if ( t1 < t2 )
437: return !ord_o ? 1 : -1;
438: }
439: }
440: return 0;
441: }
442:
1.96 noro 443: int ndl_matrix_compare(UINT *d1,UINT *d2)
444: {
445: int i,j,s;
446: int *v;
447:
448: for ( j = 0; j < nd_nvar; j++ )
449: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
450: for ( i = 0; i < nd_matrix_len; i++ ) {
451: v = nd_matrix[i];
452: for ( j = 0, s = 0; j < nd_nvar; j++ )
453: s += v[j]*nd_work_vector[j];
454: if ( s > 0 ) return 1;
455: else if ( s < 0 ) return -1;
456: }
457: return 0;
458: }
459:
1.97 noro 460: int ndl_composite_compare(UINT *d1,UINT *d2)
461: {
462: int i,j,s,start,end,len,o;
463: int *v;
464: struct sparse_weight *sw;
465:
466: for ( j = 0; j < nd_nvar; j++ )
467: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
468: for ( i = 0; i < nd_worb_len; i++ ) {
469: len = nd_worb[i].length;
470: switch ( nd_worb[i].type ) {
471: case IS_DENSE_WEIGHT:
472: v = nd_worb[i].body.dense_weight;
473: for ( j = 0, s = 0; j < len; j++ )
474: s += v[j]*nd_work_vector[j];
475: if ( s > 0 ) return 1;
476: else if ( s < 0 ) return -1;
477: break;
478: case IS_SPARSE_WEIGHT:
479: sw = nd_worb[i].body.sparse_weight;
480: for ( j = 0, s = 0; j < len; j++ )
481: s += sw[j].value*nd_work_vector[sw[j].pos];
482: if ( s > 0 ) return 1;
483: else if ( s < 0 ) return -1;
484: break;
485: case IS_BLOCK:
486: o = nd_worb[i].body.block.order;
487: start = nd_worb[i].body.block.start;
488: switch ( o ) {
489: case 0:
490: end = start+len;
491: for ( j = start, s = 0; j < end; j++ )
492: s += MUL_WEIGHT(nd_work_vector[j],j);
493: if ( s > 0 ) return 1;
494: else if ( s < 0 ) return -1;
495: for ( j = end-1; j >= start; j-- )
496: if ( nd_work_vector[j] < 0 ) return 1;
497: else if ( nd_work_vector[j] > 0 ) return -1;
498: break;
499: case 1:
500: end = start+len;
501: for ( j = start, s = 0; j < end; j++ )
502: s += MUL_WEIGHT(nd_work_vector[j],j);
503: if ( s > 0 ) return 1;
504: else if ( s < 0 ) return -1;
505: for ( j = start; j < end; j++ )
506: if ( nd_work_vector[j] > 0 ) return 1;
507: else if ( nd_work_vector[j] < 0 ) return -1;
508: break;
509: case 2:
510: for ( j = start; j < end; j++ )
511: if ( nd_work_vector[j] > 0 ) return 1;
512: else if ( nd_work_vector[j] < 0 ) return -1;
513: break;
514: }
515: break;
516: }
517: }
518: return 0;
519: }
520:
1.58 noro 521: /* TDH -> WW -> TD-> RL */
522:
1.61 noro 523: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 524: {
525: int i,m,e1,e2;
526:
527: if ( TD(d1) > TD(d2) ) return 1;
528: else if ( TD(d1) < TD(d2) ) return -1;
529: m = nd_nvar>>1;
530: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
531: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
532: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
533: }
534: if ( e1 > e2 ) return 1;
535: else if ( e1 < e2 ) return -1;
536: return ndl_lex_compare(d1,d2);
537: }
538:
1.61 noro 539: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 540: {
541: int i;
542:
1.81 noro 543: switch ( nd_wpd ) {
544: case 2:
545: if ( TD(d2) != TD(d1) ) return 0;
546: if ( d2[1] != d1[1] ) return 0;
547: return 1;
548: break;
549: case 3:
550: if ( TD(d2) != TD(d1) ) return 0;
551: if ( d2[1] != d1[1] ) return 0;
552: if ( d2[2] != d1[2] ) return 0;
553: return 1;
554: break;
555: default:
556: for ( i = 0; i < nd_wpd; i++ )
557: if ( *d1++ != *d2++ ) return 0;
558: return 1;
559: break;
560: }
1.1 noro 561: }
562:
1.61 noro 563: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 564: {
565: int i;
566:
567: switch ( nd_wpd ) {
1.41 noro 568: case 2:
1.34 noro 569: TD(d2) = TD(d1);
570: d2[1] = d1[1];
1.6 noro 571: break;
1.41 noro 572: case 3:
1.34 noro 573: TD(d2) = TD(d1);
1.6 noro 574: d2[1] = d1[1];
1.34 noro 575: d2[2] = d1[2];
1.6 noro 576: break;
577: default:
1.41 noro 578: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 579: d2[i] = d1[i];
580: break;
581: }
582: }
583:
1.61 noro 584: INLINE void ndl_zero(UINT *d)
585: {
586: int i;
587: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
588: }
589:
590: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 591: {
592: int i;
593:
1.43 noro 594: #if 1
1.6 noro 595: switch ( nd_wpd ) {
1.41 noro 596: case 2:
597: TD(d) = TD(d1)+TD(d2);
1.34 noro 598: d[1] = d1[1]+d2[1];
1.6 noro 599: break;
1.41 noro 600: case 3:
601: TD(d) = TD(d1)+TD(d2);
1.6 noro 602: d[1] = d1[1]+d2[1];
1.34 noro 603: d[2] = d1[2]+d2[2];
1.6 noro 604: break;
605: default:
1.43 noro 606: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 607: break;
608: }
1.43 noro 609: #else
610: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
611: #endif
1.6 noro 612: }
613:
1.55 noro 614: /* d1 += d2 */
1.61 noro 615: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 616: {
617: int i;
618:
619: #if 1
620: switch ( nd_wpd ) {
621: case 2:
622: TD(d1) += TD(d2);
623: d1[1] += d2[1];
624: break;
625: case 3:
626: TD(d1) += TD(d2);
627: d1[1] += d2[1];
628: d1[2] += d2[2];
629: break;
630: default:
631: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
632: break;
633: }
634: #else
635: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
636: #endif
637: }
638:
1.61 noro 639: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 640: {
641: int i;
642:
1.43 noro 643: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 644: }
645:
1.61 noro 646: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 647: {
1.61 noro 648: UINT t1,t2,u,u1,u2;
1.1 noro 649: int i,j;
650:
1.65 noro 651: #if USE_UNROLL
1.1 noro 652: switch ( nd_bpe ) {
1.62 noro 653: case 3:
654: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
655: u1 = d1[i]; u2 = d2[i];
656: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
657: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
658: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
659: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
660: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
661: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
662: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
663: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
664: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
665: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
666: }
667: return 1;
668: break;
1.1 noro 669: case 4:
1.41 noro 670: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 671: u1 = d1[i]; u2 = d2[i];
672: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 673: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
674: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
675: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
676: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
677: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
678: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
679: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 680: }
681: return 1;
682: break;
683: case 6:
1.41 noro 684: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 685: u1 = d1[i]; u2 = d2[i];
686: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 687: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
688: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
689: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
690: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 691: }
692: return 1;
693: break;
694: case 8:
1.41 noro 695: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 696: u1 = d1[i]; u2 = d2[i];
697: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 698: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
699: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
700: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 701: }
702: return 1;
703: break;
704: case 16:
1.41 noro 705: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 706: u1 = d1[i]; u2 = d2[i];
707: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 708: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 709: }
710: return 1;
711: break;
712: case 32:
1.41 noro 713: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 714: if ( d1[i] && d2[i] ) return 0;
715: return 1;
716: break;
717: default:
1.41 noro 718: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 719: u1 = d1[i]; u2 = d2[i];
720: for ( j = 0; j < nd_epw; j++ ) {
721: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
722: u1 >>= nd_bpe; u2 >>= nd_bpe;
723: }
724: }
725: return 1;
726: break;
727: }
1.65 noro 728: #else
729: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
730: u1 = d1[i]; u2 = d2[i];
731: for ( j = 0; j < nd_epw; j++ ) {
732: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
733: u1 >>= nd_bpe; u2 >>= nd_bpe;
734: }
735: }
736: return 1;
737: #endif
1.1 noro 738: }
739:
1.114 noro 740: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 741: {
1.61 noro 742: UINT u2;
1.5 noro 743: int i,j,ind,k;
1.1 noro 744:
1.5 noro 745: ind = 0;
1.65 noro 746: #if USE_UNROLL
1.5 noro 747: switch ( nd_bpe ) {
1.62 noro 748: case 3:
749: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
750: u2 = d2[i];
751: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
752: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
753: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
754: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
755: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
756: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
757: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
758: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
759: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
760: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
761: }
762: return 0;
763: break;
1.5 noro 764: case 4:
1.41 noro 765: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 766: u2 = d2[i];
767: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
768: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
769: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
770: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
771: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
772: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
773: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
774: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
775: }
776: return 0;
777: break;
778: case 6:
1.41 noro 779: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 780: u2 = d2[i];
781: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
782: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
783: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
784: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
785: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
786: }
787: return 0;
788: break;
789: case 8:
1.41 noro 790: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 791: u2 = d2[i];
792: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
793: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
794: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
795: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
796: }
797: return 0;
798: break;
799: case 16:
1.41 noro 800: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 801: u2 = d2[i];
802: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
803: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
804: }
805: return 0;
806: break;
807: case 32:
1.41 noro 808: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 809: if ( d1[i]+d2[i]<d1[i] ) return 1;
810: return 0;
811: break;
812: default:
1.41 noro 813: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 814: u2 = d2[i];
815: k = (nd_epw-1)*nd_bpe;
816: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
817: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
818: }
819: return 0;
820: break;
821: }
1.65 noro 822: #else
823: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
824: u2 = d2[i];
825: k = (nd_epw-1)*nd_bpe;
826: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
827: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
828: }
829: return 0;
830: #endif
1.1 noro 831: }
832:
1.114 noro 833: int ndl_check_bound2(int index,UINT *d2)
834: {
835: return ndl_check_bound(nd_bound[index],d2);
836: }
837:
1.61 noro 838: INLINE int ndl_hash_value(UINT *d)
1.1 noro 839: {
840: int i;
841: int r;
842:
1.34 noro 843: r = 0;
1.41 noro 844: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 845: r = ((r<<16)+d[i])%REDTAB_LEN;
846: return r;
847: }
848:
1.63 noro 849: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 850: {
1.13 noro 851: RHist r;
1.6 noro 852: int d,k,i;
1.1 noro 853:
1.63 noro 854: d = ndl_hash_value(dg);
1.13 noro 855: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 856: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 857: if ( k > 0 ) nd_notfirst++;
858: nd_found++;
1.13 noro 859: return r->index;
1.1 noro 860: }
861: }
1.13 noro 862: if ( Reverse )
863: for ( i = nd_psn-1; i >= 0; i-- ) {
864: r = nd_psh[i];
1.34 noro 865: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 866: nd_create++;
1.34 noro 867: nd_append_red(dg,i);
1.13 noro 868: return i;
869: }
870: }
871: else
872: for ( i = 0; i < nd_psn; i++ ) {
873: r = nd_psh[i];
1.34 noro 874: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 875: nd_create++;
1.34 noro 876: nd_append_red(dg,i);
1.13 noro 877: return i;
878: }
1.1 noro 879: }
1.6 noro 880: return -1;
1.1 noro 881: }
882:
1.63 noro 883: ND nd_merge(ND p1,ND p2)
884: {
885: int n,c;
886: int t,can,td1,td2;
887: ND r;
888: NM m1,m2,mr0,mr,s;
889:
890: if ( !p1 ) return p2;
891: else if ( !p2 ) return p1;
892: else {
893: can = 0;
894: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
895: c = DL_COMPARE(DL(m1),DL(m2));
896: switch ( c ) {
897: case 0:
898: s = m1; m1 = NEXT(m1);
899: can++; NEXTNM2(mr0,mr,s);
900: s = m2; m2 = NEXT(m2); FREENM(s);
901: break;
902: case 1:
903: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
904: break;
905: case -1:
906: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
907: break;
908: }
909: }
910: if ( !mr0 )
911: if ( m1 ) mr0 = m1;
912: else if ( m2 ) mr0 = m2;
913: else return 0;
914: else if ( m1 ) NEXT(mr) = m1;
915: else if ( m2 ) NEXT(mr) = m2;
916: else NEXT(mr) = 0;
917: BDY(p1) = mr0;
918: SG(p1) = MAX(SG(p1),SG(p2));
919: LEN(p1) = LEN(p1)+LEN(p2)-can;
920: FREEND(p2);
921: return p1;
922: }
923: }
924:
1.31 noro 925: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 926: {
927: int n,c;
1.34 noro 928: int t,can,td1,td2;
1.1 noro 929: ND r;
930: NM m1,m2,mr0,mr,s;
931:
1.34 noro 932: if ( !p1 ) return p2;
933: else if ( !p2 ) return p1;
1.71 noro 934: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.113 noro 935: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 936: else {
1.30 noro 937: can = 0;
1.1 noro 938: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 939: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 940: switch ( c ) {
941: case 0:
1.19 noro 942: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 943: if ( t < 0 ) t += mod;
1.1 noro 944: s = m1; m1 = NEXT(m1);
945: if ( t ) {
1.34 noro 946: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 947: } else {
1.34 noro 948: can += 2; FREENM(s);
1.17 noro 949: }
950: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 951: break;
952: case 1:
953: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
954: break;
955: case -1:
956: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
957: break;
958: }
959: }
960: if ( !mr0 )
961: if ( m1 ) mr0 = m1;
962: else if ( m2 ) mr0 = m2;
963: else return 0;
964: else if ( m1 ) NEXT(mr) = m1;
965: else if ( m2 ) NEXT(mr) = m2;
966: else NEXT(mr) = 0;
967: BDY(p1) = mr0;
968: SG(p1) = MAX(SG(p1),SG(p2));
969: LEN(p1) = LEN(p1)+LEN(p2)-can;
970: FREEND(p2);
971: return p1;
972: }
973: }
974:
975: /* XXX on opteron, the inlined manipulation of destructive additon of
976: * two NM seems to make gcc optimizer get confused, so the part is
977: * done in a function.
978: */
979:
1.113 noro 980: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 981: {
982: NM s;
1.146 noro 983: P t;
1.95 noro 984: int can;
985:
1.146 noro 986: addp(nd_vc,CP(*m1),CP(*m2),&t);
1.95 noro 987: s = *m1; *m1 = NEXT(*m1);
988: if ( t ) {
1.146 noro 989: can = 1; NEXTNM2(*mr0,*mr,s); CP(*mr) = (t);
1.95 noro 990: } else {
991: can = 2; FREENM(s);
992: }
993: s = *m2; *m2 = NEXT(*m2); FREENM(s);
994: return can;
995: }
996:
1.113 noro 997: ND nd_add_q(ND p1,ND p2)
1.95 noro 998: {
999: int n,c,can;
1000: ND r;
1001: NM m1,m2,mr0,mr,s;
1.146 noro 1002: P t;
1.95 noro 1003:
1004: if ( !p1 ) return p2;
1005: else if ( !p2 ) return p1;
1006: else {
1007: can = 0;
1008: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1009: c = DL_COMPARE(DL(m1),DL(m2));
1010: switch ( c ) {
1011: case 0:
1012: #if defined(__x86_64__)
1.113 noro 1013: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1014: #else
1.146 noro 1015: addp(nd_vc,CP(m1),CP(m2),&t);
1.95 noro 1016: s = m1; m1 = NEXT(m1);
1017: if ( t ) {
1.146 noro 1018: can++; NEXTNM2(mr0,mr,s); CP(mr) = (t);
1.95 noro 1019: } else {
1020: can += 2; FREENM(s);
1021: }
1022: s = m2; m2 = NEXT(m2); FREENM(s);
1023: #endif
1.17 noro 1024: break;
1025: case 1:
1026: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1027: break;
1028: case -1:
1029: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1030: break;
1031: }
1032: }
1033: if ( !mr0 )
1.34 noro 1034: if ( m1 ) mr0 = m1;
1035: else if ( m2 ) mr0 = m2;
1036: else return 0;
1037: else if ( m1 ) NEXT(mr) = m1;
1038: else if ( m2 ) NEXT(mr) = m2;
1039: else NEXT(mr) = 0;
1.17 noro 1040: BDY(p1) = mr0;
1041: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1042: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1043: FREEND(p2);
1044: return p1;
1045: }
1046: }
1047:
1.71 noro 1048: ND nd_add_sf(ND p1,ND p2)
1049: {
1050: int n,c,can;
1051: ND r;
1052: NM m1,m2,mr0,mr,s;
1053: int t;
1054:
1055: if ( !p1 ) return p2;
1056: else if ( !p2 ) return p1;
1057: else {
1058: can = 0;
1059: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1060: c = DL_COMPARE(DL(m1),DL(m2));
1061: switch ( c ) {
1062: case 0:
1063: t = _addsf(CM(m1),CM(m2));
1064: s = m1; m1 = NEXT(m1);
1065: if ( t ) {
1066: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1067: } else {
1068: can += 2; FREENM(s);
1069: }
1070: s = m2; m2 = NEXT(m2); FREENM(s);
1071: break;
1072: case 1:
1073: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1074: break;
1075: case -1:
1076: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1077: break;
1078: }
1079: }
1080: if ( !mr0 )
1081: if ( m1 ) mr0 = m1;
1082: else if ( m2 ) mr0 = m2;
1083: else return 0;
1084: else if ( m1 ) NEXT(mr) = m1;
1085: else if ( m2 ) NEXT(mr) = m2;
1086: else NEXT(mr) = 0;
1087: BDY(p1) = mr0;
1088: SG(p1) = MAX(SG(p1),SG(p2));
1089: LEN(p1) = LEN(p1)+LEN(p2)-can;
1090: FREEND(p2);
1091: return p1;
1092: }
1093: }
1094:
1.146 noro 1095: ND nd_reduce2(int mod,ND d,ND g,NDV p,NM mul,NDC dn)
1096: {
1097: int c,c1,c2;
1098: Q cg,cred,gcd,tq;
1099: P cgp,credp,gcdp;
1100: Obj tr,tr1;
1101:
1102: if ( mod == -1 )
1103: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1104: else if ( mod ) {
1105: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1106: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1107: } else if ( nd_vc ) {
1108: ezgcdpz(nd_vc,HCP(g),HCP(p),&gcdp);
1109: divsp(nd_vc,HCP(g),gcdp,&cgp); divsp(nd_vc,HCP(p),gcdp,&credp);
1110: chsgnp(cgp,&CP(mul));
1111: nd_mul_c_q(d,credp); nd_mul_c_q(g,credp);
1112: if ( dn ) {
1113: mulr(nd_vc,(Obj)dn->r,(Obj)credp,&tr);
1114: reductr(nd_vc,tr,&tr1); dn->r = (R)tr1;
1115: }
1116: } else {
1117: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1118: chsgnq(cg,&CQ(mul));
1119: nd_mul_c_q(d,(P)cred); nd_mul_c_q(g,(P)cred);
1120: if ( dn ) {
1121: mulq(dn->z,cred,&tq); dn->z = tq;
1122: }
1123: }
1124: return nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1125: }
1126:
1.1 noro 1127: /* ret=1 : success, ret=0 : overflow */
1.146 noro 1128: int nd_nf(int mod,ND d,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1129: {
1130: NM m,mrd,tail;
1.7 noro 1131: NM mul;
1.10 noro 1132: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1133: int c,c1,c2,dummy;
1.17 noro 1134: RHist h;
1.11 noro 1135: NDV p,red;
1.113 noro 1136: Q cg,cred,gcd,tq,qq;
1.21 noro 1137: double hmag;
1.146 noro 1138: P tp,tp1;
1139: Obj tr,tr1;
1.1 noro 1140:
1.113 noro 1141: if ( dn ) {
1142: if ( mod )
1143: dn->m = 1;
1.146 noro 1144: else if ( nd_vc )
1145: dn->r = (R)ONE;
1.113 noro 1146: else
1147: dn->z = ONE;
1148: }
1.1 noro 1149: if ( !g ) {
1.146 noro 1150: *rp = d;
1.1 noro 1151: return 1;
1152: }
1.146 noro 1153: if ( !mod ) hmag = ((double)p_mag(HCP(g)))*nd_scale;
1.21 noro 1154:
1.14 noro 1155: sugar0 = sugar = SG(g);
1.1 noro 1156: n = NV(g);
1.61 noro 1157: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.146 noro 1158: if ( d )
1159: for ( tail = BDY(d); NEXT(tail); tail = NEXT(tail) );
1160: for ( ; g; ) {
1.63 noro 1161: index = ndl_find_reducer(HDL(g));
1.6 noro 1162: if ( index >= 0 ) {
1.17 noro 1163: h = nd_psh[index];
1164: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1165: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1166: nd_free(g); nd_free(d);
1167: return 0;
1168: }
1.146 noro 1169: p = nd_demand ? ndv_load(index) : ps[index];
1170: g = nd_reduce2(mod,d,g,p,mul,dn);
1.34 noro 1171: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.146 noro 1172: if ( !mod && g && ((double)(p_mag(HCP(g))) > hmag) ) {
1173: tp = HCP(g);
1.21 noro 1174: nd_removecont2(d,g);
1.69 noro 1175: if ( dn ) {
1.146 noro 1176: if ( nd_vc ) {
1177: divsp(nd_vc,tp,HCP(g),&tp1);
1178: divr(nd_vc,(Obj)dn->r,(Obj)tp1,&tr);
1179: reductr(nd_vc,(Obj)tr,&tr1); dn->r = (R)tr1;
1180: } else {
1181: divq((Q)tp,HCQ(g),&qq);
1182: divq(dn->z,qq,&tq); dn->z = tq;
1183: }
1.69 noro 1184: }
1.146 noro 1185: hmag = ((double)p_mag(HCP(g)))*nd_scale;
1.21 noro 1186: }
1.1 noro 1187: } else if ( !full ) {
1188: *rp = g;
1189: return 1;
1190: } else {
1191: m = BDY(g);
1192: if ( NEXT(m) ) {
1.34 noro 1193: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1194: } else {
1195: FREEND(g); g = 0;
1196: }
1197: if ( d ) {
1.34 noro 1198: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1199: } else {
1.34 noro 1200: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1201: }
1202: }
1203: }
1.34 noro 1204: if ( d ) SG(d) = sugar;
1.1 noro 1205: *rp = d;
1206: return 1;
1207: }
1.28 noro 1208:
1.53 noro 1209: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1210: {
1211: int hindex,index;
1212: NDV p;
1213: ND u,d,red;
1214: NODE l;
1.31 noro 1215: NM mul,m,mrd,tail;
1.25 noro 1216: int sugar,psugar,n,h_reducible;
1217: PGeoBucket bucket;
1218: int c,c1,c2;
1.113 noro 1219: Q cg,cred,gcd,zzz;
1.25 noro 1220: RHist h;
1.28 noro 1221: double hmag,gmag;
1.77 noro 1222: int count = 0;
1223: int hcount = 0;
1.25 noro 1224:
1225: if ( !g ) {
1226: *rp = 0;
1227: return 1;
1228: }
1229: sugar = SG(g);
1230: n = NV(g);
1.113 noro 1231: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1232: bucket = create_pbucket();
1.31 noro 1233: add_pbucket(mod,bucket,g);
1.25 noro 1234: d = 0;
1.61 noro 1235: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1236: while ( 1 ) {
1.113 noro 1237: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1238: if ( hindex < 0 ) {
1.77 noro 1239: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1240: if ( d ) SG(d) = sugar;
1.25 noro 1241: *rp = d;
1242: return 1;
1243: }
1244: g = bucket->body[hindex];
1.63 noro 1245: index = ndl_find_reducer(HDL(g));
1.25 noro 1246: if ( index >= 0 ) {
1.77 noro 1247: count++;
1248: if ( !d ) hcount++;
1.25 noro 1249: h = nd_psh[index];
1250: ndl_sub(HDL(g),DL(h),DL(mul));
1251: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1252: nd_free(d);
1.25 noro 1253: free_pbucket(bucket);
1254: *rp = 0;
1255: return 0;
1256: }
1.53 noro 1257: p = ps[index];
1.71 noro 1258: if ( mod == -1 )
1259: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1260: else if ( mod ) {
1.25 noro 1261: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1262: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1263: } else {
1.113 noro 1264: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1265: chsgnq(cg,&CQ(mul));
1.146 noro 1266: nd_mul_c_q(d,(P)cred);
1.113 noro 1267: mulq_pbucket(bucket,cred);
1.26 noro 1268: g = bucket->body[hindex];
1.113 noro 1269: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1270: }
1.55 noro 1271: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1272: bucket->body[hindex] = nd_remove_head(g);
1273: red = nd_remove_head(red);
1.31 noro 1274: add_pbucket(mod,bucket,red);
1.34 noro 1275: psugar = SG(p)+TD(DL(mul));
1276: sugar = MAX(sugar,psugar);
1.28 noro 1277: if ( !mod && hmag && (gmag > hmag) ) {
1278: g = normalize_pbucket(mod,bucket);
1279: if ( !g ) {
1.34 noro 1280: if ( d ) SG(d) = sugar;
1.28 noro 1281: *rp = d;
1282: return 1;
1283: }
1284: nd_removecont2(d,g);
1.113 noro 1285: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1286: add_pbucket(mod,bucket,g);
1.28 noro 1287: }
1.25 noro 1288: } else if ( !full ) {
1289: g = normalize_pbucket(mod,bucket);
1.34 noro 1290: if ( g ) SG(g) = sugar;
1.25 noro 1291: *rp = g;
1292: return 1;
1293: } else {
1294: m = BDY(g);
1295: if ( NEXT(m) ) {
1.34 noro 1296: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1297: } else {
1298: FREEND(g); g = 0;
1299: }
1300: bucket->body[hindex] = g;
1301: NEXT(m) = 0;
1302: if ( d ) {
1.34 noro 1303: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1304: } else {
1.34 noro 1305: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1306: }
1307: }
1308: }
1309: }
1.27 noro 1310:
1.61 noro 1311: /* input : list of NDV, cand : list of NDV */
1.28 noro 1312:
1.61 noro 1313: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1314: {
1315: int n,i,stat;
1316: ND nf,d;
1.61 noro 1317: NDV r;
1.45 noro 1318: NODE t,s;
1.113 noro 1319: union oNDC dn;
1.45 noro 1320:
1.128 noro 1321: ndv_setup(0,0,cand,0);
1.31 noro 1322: n = length(cand);
1.28 noro 1323:
1324: /* membercheck : list is a subset of Id(cand) ? */
1325: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1326: again:
1.61 noro 1327: if ( nd_bpe > obpe )
1328: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1329: else
1330: r = (NDV)BDY(t);
1331: d = ndvtond(0,r);
1.146 noro 1332: stat = nd_nf(0,0,d,nd_ps,0,0,&nf);
1.45 noro 1333: if ( !stat ) {
1.103 noro 1334: nd_reconstruct(0,0);
1.45 noro 1335: goto again;
1336: } else if ( nf ) return 0;
1.71 noro 1337: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1338: }
1.71 noro 1339: if ( DP_Print ) { printf("\n"); }
1.28 noro 1340: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1341: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1342: /* XXX */
1.23 noro 1343: return 1;
1344: }
1.1 noro 1345:
1346: ND nd_remove_head(ND p)
1347: {
1348: NM m;
1349:
1350: m = BDY(p);
1351: if ( !NEXT(m) ) {
1.34 noro 1352: FREEND(p); p = 0;
1.31 noro 1353: } else {
1.34 noro 1354: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1355: }
1.1 noro 1356: FREENM(m);
1357: return p;
1358: }
1359:
1.69 noro 1360: ND nd_separate_head(ND p,ND *head)
1361: {
1362: NM m,m0;
1363: ND r;
1364:
1365: m = BDY(p);
1366: if ( !NEXT(m) ) {
1367: *head = p; p = 0;
1368: } else {
1369: m0 = m;
1370: BDY(p) = NEXT(m); LEN(p)--;
1371: NEXT(m0) = 0;
1372: MKND(NV(p),m0,1,r);
1373: *head = r;
1374: }
1375: return p;
1376: }
1377:
1.1 noro 1378: PGeoBucket create_pbucket()
1379: {
1380: PGeoBucket g;
1381:
1382: g = CALLOC(1,sizeof(struct oPGeoBucket));
1383: g->m = -1;
1384: return g;
1385: }
1386:
1.25 noro 1387: void free_pbucket(PGeoBucket b) {
1388: int i;
1389:
1.26 noro 1390: for ( i = 0; i <= b->m; i++ )
1.25 noro 1391: if ( b->body[i] ) {
1392: nd_free(b->body[i]);
1393: b->body[i] = 0;
1394: }
1395: GC_free(b);
1396: }
1397:
1.63 noro 1398: void add_pbucket_symbolic(PGeoBucket g,ND d)
1399: {
1400: int l,i,k,m;
1401:
1402: if ( !d )
1403: return;
1404: l = LEN(d);
1405: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1406: /* 2^(k-1) < l <= 2^k (=m) */
1407: d = nd_merge(g->body[k],d);
1408: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1409: g->body[k] = 0;
1410: d = nd_merge(g->body[k+1],d);
1411: }
1412: g->body[k] = d;
1413: g->m = MAX(g->m,k);
1414: }
1415:
1.31 noro 1416: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1417: {
1.31 noro 1418: int l,i,k,m;
1.1 noro 1419:
1.31 noro 1420: if ( !d )
1421: return;
1422: l = LEN(d);
1.29 noro 1423: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1424: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1425: d = nd_add(mod,g->body[k],d);
1426: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1427: g->body[k] = 0;
1.31 noro 1428: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1429: }
1430: g->body[k] = d;
1431: g->m = MAX(g->m,k);
1432: }
1433:
1.113 noro 1434: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1435: {
1436: int k;
1437:
1438: for ( k = 0; k <= g->m; k++ )
1.146 noro 1439: nd_mul_c_q(g->body[k],(P)c);
1.26 noro 1440: }
1441:
1.63 noro 1442: NM remove_head_pbucket_symbolic(PGeoBucket g)
1443: {
1444: int j,i,k,c;
1445: NM head;
1446:
1447: k = g->m;
1448: j = -1;
1449: for ( i = 0; i <= k; i++ ) {
1450: if ( !g->body[i] ) continue;
1451: if ( j < 0 ) j = i;
1452: else {
1453: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1454: if ( c > 0 )
1455: j = i;
1456: else if ( c == 0 )
1457: g->body[i] = nd_remove_head(g->body[i]);
1458: }
1459: }
1460: if ( j < 0 ) return 0;
1461: else {
1462: head = BDY(g->body[j]);
1463: if ( !NEXT(head) ) {
1464: FREEND(g->body[j]);
1465: g->body[j] = 0;
1466: } else {
1467: BDY(g->body[j]) = NEXT(head);
1468: LEN(g->body[j])--;
1469: }
1470: return head;
1471: }
1472: }
1473:
1.19 noro 1474: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1475: {
1476: int j,i,c,k,nv,sum;
1.61 noro 1477: UINT *di,*dj;
1.1 noro 1478: ND gi,gj;
1479:
1480: k = g->m;
1481: while ( 1 ) {
1482: j = -1;
1483: for ( i = 0; i <= k; i++ ) {
1484: if ( !(gi = g->body[i]) )
1485: continue;
1486: if ( j < 0 ) {
1487: j = i;
1488: gj = g->body[j];
1489: dj = HDL(gj);
1.14 noro 1490: sum = HCM(gj);
1.1 noro 1491: } else {
1.34 noro 1492: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1493: if ( c > 0 ) {
1.34 noro 1494: if ( sum ) HCM(gj) = sum;
1495: else g->body[j] = nd_remove_head(gj);
1.1 noro 1496: j = i;
1497: gj = g->body[j];
1498: dj = HDL(gj);
1.14 noro 1499: sum = HCM(gj);
1.1 noro 1500: } else if ( c == 0 ) {
1.71 noro 1501: if ( mod == -1 )
1502: sum = _addsf(sum,HCM(gi));
1503: else {
1504: sum = sum+HCM(gi)-mod;
1505: if ( sum < 0 ) sum += mod;
1506: }
1.1 noro 1507: g->body[i] = nd_remove_head(gi);
1508: }
1509: }
1510: }
1.34 noro 1511: if ( j < 0 ) return -1;
1.1 noro 1512: else if ( sum ) {
1.14 noro 1513: HCM(gj) = sum;
1.26 noro 1514: return j;
1.31 noro 1515: } else
1.26 noro 1516: g->body[j] = nd_remove_head(gj);
1517: }
1518: }
1519:
1.113 noro 1520: int head_pbucket_q(PGeoBucket g)
1.26 noro 1521: {
1522: int j,i,c,k,nv;
1.113 noro 1523: Q sum,t;
1.26 noro 1524: ND gi,gj;
1525:
1526: k = g->m;
1527: while ( 1 ) {
1528: j = -1;
1529: for ( i = 0; i <= k; i++ ) {
1.34 noro 1530: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1531: if ( j < 0 ) {
1532: j = i;
1533: gj = g->body[j];
1.113 noro 1534: sum = HCQ(gj);
1.26 noro 1535: } else {
1536: nv = NV(gi);
1.34 noro 1537: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1538: if ( c > 0 ) {
1.113 noro 1539: if ( sum ) HCQ(gj) = sum;
1.34 noro 1540: else g->body[j] = nd_remove_head(gj);
1.26 noro 1541: j = i;
1542: gj = g->body[j];
1.113 noro 1543: sum = HCQ(gj);
1.26 noro 1544: } else if ( c == 0 ) {
1.113 noro 1545: addq(sum,HCQ(gi),&t);
1546: sum = t;
1.26 noro 1547: g->body[i] = nd_remove_head(gi);
1548: }
1549: }
1550: }
1.34 noro 1551: if ( j < 0 ) return -1;
1.26 noro 1552: else if ( sum ) {
1.113 noro 1553: HCQ(gj) = sum;
1.1 noro 1554: return j;
1.31 noro 1555: } else
1.1 noro 1556: g->body[j] = nd_remove_head(gj);
1557: }
1558: }
1559:
1.25 noro 1560: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1561: {
1.31 noro 1562: int i;
1.1 noro 1563: ND r,t;
1564:
1565: r = 0;
1.28 noro 1566: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1567: r = nd_add(mod,r,g->body[i]);
1.28 noro 1568: g->body[i] = 0;
1569: }
1570: g->m = -1;
1.1 noro 1571: return r;
1572: }
1573:
1.122 noro 1574: int do_diagonalize(int sugar,int m)
1.92 noro 1575: {
1576: int i,nh,stat;
1577: NODE r,g,t;
1578: ND h,nf,s,head;
1579: NDV nfv;
1.113 noro 1580: Q q,num,den;
1.146 noro 1581: P nm,nmp,dn,mnp,dnp;
1.92 noro 1582:
1.143 noro 1583: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.92 noro 1584: if ( nd_demand )
1585: nfv = ndv_load(i);
1586: else
1587: nfv = nd_ps[i];
1588: s = ndvtond(m,nfv);
1589: s = nd_separate_head(s,&head);
1.146 noro 1590: stat = nd_nf(m,head,s,nd_ps,1,0,&nf);
1.122 noro 1591: if ( !stat ) return 0;
1.92 noro 1592: ndv_free(nfv);
1593: nd_removecont(m,nf);
1594: nfv = ndtondv(m,nf);
1595: nd_free(nf);
1596: nd_bound[i] = ndv_compute_bound(nfv);
1597: if ( nd_demand ) {
1598: ndv_save(nfv,i);
1599: ndv_free(nfv);
1600: } else
1601: nd_ps[i] = nfv;
1602: }
1.122 noro 1603: return 1;
1.92 noro 1604: }
1605:
1.27 noro 1606: /* return value = 0 => input is not a GB */
1607:
1.77 noro 1608: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1609: {
1610: int i,nh,sugar,stat;
1.23 noro 1611: NODE r,g,t;
1.1 noro 1612: ND_pairs d;
1613: ND_pairs l;
1.117 noro 1614: ND h,nf,s,head,nf1;
1.63 noro 1615: NDV nfv;
1.113 noro 1616: Q q,num,den;
1617: union oNDC dn;
1.140 noro 1618: int diag_count = 0;
1.1 noro 1619:
1.23 noro 1620: g = 0; d = 0;
1621: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1622: d = update_pairs(d,g,i);
1623: g = update_base(g,i);
1624: }
1625: sugar = 0;
1626: while ( d ) {
1627: again:
1628: l = nd_minp(d,&d);
1.14 noro 1629: if ( SG(l) != sugar ) {
1.122 noro 1630: if ( ishomo ) {
1.140 noro 1631: diag_count = 0;
1.122 noro 1632: stat = do_diagonalize(sugar,m);
1633: if ( !stat ) {
1634: NEXT(l) = d; d = l;
1635: d = nd_reconstruct(0,d);
1636: goto again;
1637: }
1638: }
1.14 noro 1639: sugar = SG(l);
1.71 noro 1640: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1641: }
1.53 noro 1642: stat = nd_sp(m,0,l,&h);
1.1 noro 1643: if ( !stat ) {
1644: NEXT(l) = d; d = l;
1.103 noro 1645: d = nd_reconstruct(0,d);
1.1 noro 1646: goto again;
1647: }
1.41 noro 1648: #if USE_GEOBUCKET
1.146 noro 1649: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1650: #else
1.146 noro 1651: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1652: #endif
1.1 noro 1653: if ( !stat ) {
1654: NEXT(l) = d; d = l;
1.103 noro 1655: d = nd_reconstruct(0,d);
1.1 noro 1656: goto again;
1657: } else if ( nf ) {
1.27 noro 1658: if ( checkonly ) return 0;
1.71 noro 1659: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1660: nd_removecont(m,nf);
1.118 noro 1661: if ( !m && nd_nalg ) {
1662: nd_monic(0,&nf);
1663: nd_removecont(m,nf);
1.117 noro 1664: }
1.63 noro 1665: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1666: nh = ndv_newps(m,nfv,0);
1.141 noro 1667: if ( !m && (ishomo && ++diag_count == diag_period) ) {
1.140 noro 1668: diag_count = 0;
1669: stat = do_diagonalize(sugar,m);
1670: if ( !stat ) {
1671: NEXT(l) = d; d = l;
1672: d = nd_reconstruct(1,d);
1673: goto again;
1674: }
1675: }
1.1 noro 1676: d = update_pairs(d,g,nh);
1677: g = update_base(g,nh);
1678: FREENDP(l);
1679: } else {
1.71 noro 1680: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1681: FREENDP(l);
1682: }
1683: }
1.77 noro 1684: if ( nd_demand )
1685: for ( t = g; t; t = NEXT(t) )
1686: BDY(t) = (pointer)ndv_load((int)BDY(t));
1687: else
1688: for ( t = g; t; t = NEXT(t) )
1689: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.149 ! noro 1690: if ( !checkonly && DP_Print ) { printf("nd_gb done.\n"); fflush(stdout); }
1.1 noro 1691: return g;
1692: }
1693:
1.122 noro 1694: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1695: {
1696: int i,nh,stat;
1697: NODE r,g,t;
1698: ND h,nf,nfq,s,head;
1699: NDV nfv,nfqv;
1.113 noro 1700: Q q,den,num;
1.91 noro 1701:
1.143 noro 1702: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1.91 noro 1703: /* for nd_ps */
1704: s = ndvtond(m,nd_ps[i]);
1705: s = nd_separate_head(s,&head);
1.122 noro 1706: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1707: if ( !stat ) return 0;
1.91 noro 1708: nf = nd_add(m,head,nf);
1709: ndv_free(nd_ps[i]);
1710: nd_ps[i] = ndtondv(m,nf);
1711: nd_free(nf);
1712:
1713: /* for nd_ps_trace */
1714: if ( nd_demand )
1715: nfv = ndv_load(i);
1716: else
1717: nfv = nd_ps_trace[i];
1718: s = ndvtond(0,nfv);
1719: s = nd_separate_head(s,&head);
1.146 noro 1720: stat = nd_nf(0,head,s,nd_ps_trace,1,0,&nf);
1.122 noro 1721: if ( !stat ) return 0;
1.91 noro 1722: ndv_free(nfv);
1723: nd_removecont(0,nf);
1724: nfv = ndtondv(0,nf);
1725: nd_free(nf);
1726: nd_bound[i] = ndv_compute_bound(nfv);
1727: if ( nd_demand ) {
1728: ndv_save(nfv,i);
1729: ndv_free(nfv);
1730: } else
1731: nd_ps_trace[i] = nfv;
1732: }
1.122 noro 1733: return 1;
1.91 noro 1734: }
1735:
1.118 noro 1736: static struct oEGT eg_invdalg;
1737: struct oEGT eg_le;
1738:
1.147 noro 1739: void nd_subst_vector(VL vl,P p,NODE subst,P *r)
1740: {
1741: NODE tn;
1742: P p1;
1743:
1744: for ( tn = subst; tn; tn = NEXT(NEXT(tn)) ) {
1745: substp(vl,p,BDY(tn),BDY(NEXT(tn)),&p1); p = p1;
1746: }
1747: *r = p;
1748: }
1749:
1.82 noro 1750: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1751: {
1752: int i,nh,sugar,stat;
1.23 noro 1753: NODE r,g,t;
1.20 noro 1754: ND_pairs d;
1755: ND_pairs l;
1.82 noro 1756: ND h,nf,nfq,s,head;
1.63 noro 1757: NDV nfv,nfqv;
1.113 noro 1758: Q q,den,num;
1.147 noro 1759: P hc;
1.113 noro 1760: union oNDC dn;
1.118 noro 1761: struct oEGT eg_monic,egm0,egm1;
1.130 noro 1762: int diag_count = 0;
1.20 noro 1763:
1.118 noro 1764: init_eg(&eg_monic);
1765: init_eg(&eg_invdalg);
1766: init_eg(&eg_le);
1.23 noro 1767: g = 0; d = 0;
1768: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1769: d = update_pairs(d,g,i);
1770: g = update_base(g,i);
1771: }
1772: sugar = 0;
1773: while ( d ) {
1774: again:
1775: l = nd_minp(d,&d);
1776: if ( SG(l) != sugar ) {
1.130 noro 1777: #if 1
1.122 noro 1778: if ( ishomo ) {
1779: stat = do_diagonalize_trace(sugar,m);
1.130 noro 1780: diag_count = 0;
1.122 noro 1781: if ( !stat ) {
1782: NEXT(l) = d; d = l;
1783: d = nd_reconstruct(1,d);
1784: goto again;
1785: }
1786: }
1.130 noro 1787: #endif
1.20 noro 1788: sugar = SG(l);
1.71 noro 1789: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1790: }
1.53 noro 1791: stat = nd_sp(m,0,l,&h);
1.20 noro 1792: if ( !stat ) {
1793: NEXT(l) = d; d = l;
1.103 noro 1794: d = nd_reconstruct(1,d);
1.20 noro 1795: goto again;
1796: }
1.41 noro 1797: #if USE_GEOBUCKET
1.53 noro 1798: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1799: #else
1.146 noro 1800: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1801: #endif
1.20 noro 1802: if ( !stat ) {
1803: NEXT(l) = d; d = l;
1.103 noro 1804: d = nd_reconstruct(1,d);
1.20 noro 1805: goto again;
1806: } else if ( nf ) {
1.77 noro 1807: if ( nd_demand ) {
1808: nfqv = ndv_load(nd_psn);
1809: nfq = ndvtond(0,nfqv);
1810: } else
1811: nfq = 0;
1812: if ( !nfq ) {
1.146 noro 1813: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,0,h,nd_ps_trace,!Top,0,&nfq) ) {
1.77 noro 1814: NEXT(l) = d; d = l;
1.103 noro 1815: d = nd_reconstruct(1,d);
1.77 noro 1816: goto again;
1817: }
1.70 noro 1818: }
1.20 noro 1819: if ( nfq ) {
1.70 noro 1820: /* m|HC(nfq) => failure */
1.147 noro 1821: if ( nd_vc ) {
1822: nd_subst_vector(nd_vc,HCP(nfq),nd_subst,&hc); q = (Q)hc;
1823: } else
1824: q = HCQ(nfq);
1825: if ( !rem(NM(q),m) ) return 0;
1.63 noro 1826:
1.71 noro 1827: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.118 noro 1828: if ( nd_nalg ) {
1829: /* m|DN(HC(nf)^(-1)) => failure */
1830: get_eg(&egm0);
1831: if ( !nd_monic(m,&nfq) ) return 0;
1832: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
1833: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1834: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
1835: } else {
1836: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1837: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1838: }
1.77 noro 1839: nh = ndv_newps(0,nfv,nfqv);
1.130 noro 1840: if ( ishomo && ++diag_count == diag_period ) {
1841: diag_count = 0;
1842: stat = do_diagonalize_trace(sugar,m);
1843: if ( !stat ) {
1844: NEXT(l) = d; d = l;
1845: d = nd_reconstruct(1,d);
1846: goto again;
1847: }
1848: }
1.20 noro 1849: d = update_pairs(d,g,nh);
1850: g = update_base(g,nh);
1851: } else {
1.71 noro 1852: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1853: }
1854: } else {
1.71 noro 1855: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1856: }
1857: FREENDP(l);
1858: }
1.77 noro 1859: if ( nd_demand )
1860: for ( t = g; t; t = NEXT(t) )
1861: BDY(t) = (pointer)ndv_load((int)BDY(t));
1862: else
1863: for ( t = g; t; t = NEXT(t) )
1864: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.118 noro 1865: if ( nd_nalg ) {
1866: print_eg("monic",&eg_monic);
1867: print_eg("invdalg",&eg_invdalg);
1868: print_eg("le",&eg_le);
1869: }
1.149 ! noro 1870: if ( DP_Print ) { printf("nd_gb_trace done.\n"); fflush(stdout); }
1.20 noro 1871: return g;
1872: }
1873:
1.23 noro 1874: int ndv_compare(NDV *p1,NDV *p2)
1875: {
1.34 noro 1876: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1877: }
1878:
1879: int ndv_compare_rev(NDV *p1,NDV *p2)
1880: {
1.34 noro 1881: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1882: }
1883:
1.61 noro 1884: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1885: {
1.69 noro 1886: int i,n,stat;
1887: ND nf,g,head;
1.23 noro 1888: NODE t,a0,a;
1.113 noro 1889: union oNDC dn;
1.69 noro 1890: NDV *w;
1.113 noro 1891: Q q,num,den;
1.23 noro 1892:
1.69 noro 1893: n = length(f);
1894: #if 0
1895: w = (NDV *)ALLOCA(n*sizeof(NDV));
1896: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1897: qsort(w,n,sizeof(NDV),
1898: (int (*)(const void *,const void *))ndv_compare);
1899: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1900: #endif
1.128 noro 1901: ndv_setup(m,0,f,0);
1.69 noro 1902: for ( i = 0; i < n; ) {
1903: g = ndvtond(m,nd_ps[i]);
1904: g = nd_separate_head(g,&head);
1.146 noro 1905: stat = nd_nf(m,head,g,nd_ps,1,0,&nf);
1.50 noro 1906: if ( !stat )
1.103 noro 1907: nd_reconstruct(0,0);
1.69 noro 1908: else {
1.71 noro 1909: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1910: ndv_free(nd_ps[i]);
1.24 noro 1911: nd_removecont(m,nf);
1.69 noro 1912: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1913: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1914: i++;
1.23 noro 1915: }
1916: }
1.71 noro 1917: if ( DP_Print ) { printf("\n"); }
1.23 noro 1918: for ( a0 = 0, i = 0; i < n; i++ ) {
1919: NEXTNODE(a0,a);
1.69 noro 1920: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1921: }
1922: NEXT(a) = 0;
1923: return a0;
1924: }
1925:
1.1 noro 1926: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1927: {
1928: ND_pairs d1,nd,cur,head,prev,remove;
1929:
1930: if ( !g ) return d;
1931: d = crit_B(d,t);
1932: d1 = nd_newpairs(g,t);
1933: d1 = crit_M(d1);
1934: d1 = crit_F(d1);
1.55 noro 1935: if ( do_weyl )
1936: head = d1;
1937: else {
1938: prev = 0; cur = head = d1;
1939: while ( cur ) {
1940: if ( crit_2( cur->i1,cur->i2 ) ) {
1941: remove = cur;
1942: if ( !prev ) head = cur = NEXT(cur);
1943: else cur = NEXT(prev) = NEXT(cur);
1944: FREENDP(remove);
1945: } else {
1946: prev = cur; cur = NEXT(cur);
1947: }
1.1 noro 1948: }
1949: }
1950: if ( !d )
1951: return head;
1952: else {
1953: nd = d;
1.34 noro 1954: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1955: NEXT(nd) = head;
1956: return d;
1957: }
1958: }
1959:
1960: ND_pairs nd_newpairs( NODE g, int t )
1961: {
1962: NODE h;
1.61 noro 1963: UINT *dl;
1.34 noro 1964: int ts,s;
1.1 noro 1965: ND_pairs r,r0;
1966:
1.20 noro 1967: dl = DL(nd_psh[t]);
1.34 noro 1968: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1969: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1970: NEXTND_pairs(r0,r);
1971: r->i1 = (int)BDY(h);
1972: r->i2 = t;
1.20 noro 1973: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1974: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1975: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1976: }
1977: NEXT(r) = 0;
1978: return r0;
1979: }
1980:
1981: ND_pairs crit_B( ND_pairs d, int s )
1982: {
1983: ND_pairs cur,head,prev,remove;
1.61 noro 1984: UINT *t,*tl,*lcm;
1.1 noro 1985: int td,tdl;
1986:
1987: if ( !d ) return 0;
1.20 noro 1988: t = DL(nd_psh[s]);
1.1 noro 1989: prev = 0;
1990: head = cur = d;
1.61 noro 1991: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1992: while ( cur ) {
1993: tl = cur->lcm;
1994: if ( ndl_reducible(tl,t)
1.20 noro 1995: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1996: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1997: remove = cur;
1998: if ( !prev ) {
1999: head = cur = NEXT(cur);
2000: } else {
2001: cur = NEXT(prev) = NEXT(cur);
2002: }
2003: FREENDP(remove);
2004: } else {
1.34 noro 2005: prev = cur; cur = NEXT(cur);
1.1 noro 2006: }
2007: }
2008: return head;
2009: }
2010:
2011: ND_pairs crit_M( ND_pairs d1 )
2012: {
2013: ND_pairs e,d2,d3,dd,p;
1.61 noro 2014: UINT *id,*jd;
1.1 noro 2015:
2016: for ( dd = 0, e = d1; e; e = d3 ) {
2017: if ( !(d2 = NEXT(e)) ) {
2018: NEXT(e) = dd;
2019: return e;
2020: }
1.34 noro 2021: id = LCM(e);
1.1 noro 2022: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2023: p = NEXT(d2);
2024: jd = LCM(d2);
2025: if ( ndl_equal(jd,id) )
2026: ;
2027: else if ( TD(jd) > TD(id) )
1.1 noro 2028: if ( ndl_reducible(jd,id) ) continue;
2029: else ;
1.34 noro 2030: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2031: NEXT(d2) = d3;
2032: d3 = d2;
2033: }
2034: NEXT(e) = dd;
2035: dd = e;
2036: continue;
2037: /**/
2038: delit: NEXT(d2) = d3;
2039: d3 = d2;
2040: for ( ; p; p = d2 ) {
2041: d2 = NEXT(p);
2042: NEXT(p) = d3;
2043: d3 = p;
2044: }
2045: FREENDP(e);
2046: }
2047: return dd;
2048: }
2049:
2050: ND_pairs crit_F( ND_pairs d1 )
2051: {
2052: ND_pairs rest, head,remove;
2053: ND_pairs last, p, r, w;
2054: int s;
2055:
2056: for ( head = last = 0, p = d1; NEXT(p); ) {
2057: r = w = equivalent_pairs(p,&rest);
1.14 noro 2058: s = SG(r);
1.1 noro 2059: w = NEXT(w);
2060: while ( w ) {
2061: if ( crit_2(w->i1,w->i2) ) {
2062: r = w;
2063: w = NEXT(w);
2064: while ( w ) {
2065: remove = w;
2066: w = NEXT(w);
2067: FREENDP(remove);
2068: }
2069: break;
1.14 noro 2070: } else if ( SG(w) < s ) {
1.1 noro 2071: FREENDP(r);
2072: r = w;
1.14 noro 2073: s = SG(r);
1.1 noro 2074: w = NEXT(w);
2075: } else {
2076: remove = w;
2077: w = NEXT(w);
2078: FREENDP(remove);
2079: }
2080: }
2081: if ( last ) NEXT(last) = r;
2082: else head = r;
2083: NEXT(last = r) = 0;
2084: p = rest;
2085: if ( !p ) return head;
2086: }
2087: if ( !last ) return p;
2088: NEXT(last) = p;
2089: return head;
2090: }
2091:
2092: int crit_2( int dp1, int dp2 )
2093: {
1.20 noro 2094: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2095: }
2096:
1.40 noro 2097: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2098: {
2099: ND_pairs w,p,r,s;
1.61 noro 2100: UINT *d;
1.1 noro 2101:
2102: w = d1;
1.34 noro 2103: d = LCM(w);
1.1 noro 2104: s = NEXT(w);
2105: NEXT(w) = 0;
2106: for ( r = 0; s; s = p ) {
2107: p = NEXT(s);
1.34 noro 2108: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2109: NEXT(s) = w; w = s;
1.1 noro 2110: } else {
1.39 noro 2111: NEXT(s) = r; r = s;
1.1 noro 2112: }
2113: }
2114: *prest = r;
2115: return w;
2116: }
2117:
2118: NODE update_base(NODE nd,int ndp)
2119: {
1.61 noro 2120: UINT *dl, *dln;
1.137 noro 2121: NODE last, p, head;
1.1 noro 2122:
1.20 noro 2123: dl = DL(nd_psh[ndp]);
1.1 noro 2124: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2125: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2126: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2127: p = NEXT(p);
2128: if ( last ) NEXT(last) = p;
2129: } else {
2130: if ( !last ) head = p;
2131: p = NEXT(last = p);
2132: }
2133: }
2134: head = append_one(head,ndp);
2135: return head;
2136: }
2137:
2138: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2139: {
2140: ND_pairs m,ml,p,l;
1.61 noro 2141: UINT *lcm;
1.33 noro 2142: int s,td,len,tlen,c,c1;
1.1 noro 2143:
2144: if ( !(p = NEXT(m = d)) ) {
2145: *prest = p;
2146: NEXT(m) = 0;
2147: return m;
2148: }
1.14 noro 2149: s = SG(m);
1.77 noro 2150: if ( !NoSugar ) {
2151: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2152: if ( (SG(p) < s)
2153: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2154: ml = l; m = p; s = SG(m);
2155: }
2156: } else {
2157: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2158: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2159: ml = l; m = p; s = SG(m);
2160: }
2161: }
1.1 noro 2162: if ( !ml ) *prest = NEXT(m);
2163: else {
2164: NEXT(ml) = NEXT(m);
2165: *prest = d;
2166: }
2167: NEXT(m) = 0;
2168: return m;
2169: }
2170:
1.63 noro 2171: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2172: {
1.69 noro 2173: int msugar,i;
1.63 noro 2174: ND_pairs t,dm0,dm,dr0,dr;
2175:
2176: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2177: if ( SG(t) < msugar ) msugar = SG(t);
2178: dm0 = 0; dr0 = 0;
1.69 noro 2179: for ( i = 0, t = d; t; t = NEXT(t) )
2180: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2181: if ( dm0 ) NEXT(dm) = t;
2182: else dm0 = t;
2183: dm = t;
1.69 noro 2184: i++;
1.63 noro 2185: } else {
2186: if ( dr0 ) NEXT(dr) = t;
2187: else dr0 = t;
2188: dr = t;
2189: }
2190: NEXT(dm) = 0;
2191: if ( dr0 ) NEXT(dr) = 0;
2192: *prest = dr0;
2193: return dm0;
2194: }
2195:
1.77 noro 2196: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2197: {
1.3 noro 2198: int len;
1.13 noro 2199: RHist r;
1.20 noro 2200: NDV b;
1.3 noro 2201:
1.1 noro 2202: if ( nd_psn == nd_pslen ) {
2203: nd_pslen *= 2;
1.11 noro 2204: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2205: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2206: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2207: nd_bound = (UINT **)
2208: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2209: }
1.39 noro 2210: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2211: nd_ps[nd_psn] = a;
1.39 noro 2212: if ( aq ) {
1.63 noro 2213: nd_ps_trace[nd_psn] = aq;
2214: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2215: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2216: } else {
1.63 noro 2217: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2218: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2219: }
1.77 noro 2220: if ( nd_demand ) {
2221: if ( aq ) {
2222: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2223: nd_ps_trace[nd_psn] = 0;
2224: } else {
2225: ndv_save(nd_ps[nd_psn],nd_psn);
2226: nd_ps[nd_psn] = 0;
2227: }
2228: }
1.1 noro 2229: return nd_psn++;
2230: }
2231:
1.128 noro 2232: void ndv_setup(int mod,int trace,NODE f,int dont_sort)
1.1 noro 2233: {
1.5 noro 2234: int i,j,td,len,max;
1.1 noro 2235: NODE s,s0,f0;
1.61 noro 2236: UINT *d;
1.13 noro 2237: RHist r;
1.69 noro 2238: NDV *w;
1.61 noro 2239: NDV a,am;
1.11 noro 2240:
2241: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2242:
1.69 noro 2243: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2244: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2245: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
1.139 noro 2246: if ( !dont_sort ) {
2247: /* XXX heuristic */
2248: if ( !nd_ord->id && (nd_ord->ord.simple<2) )
2249: qsort(w,nd_psn,sizeof(NDV),
2250: (int (*)(const void *,const void *))ndv_compare_rev);
2251: else
2252: qsort(w,nd_psn,sizeof(NDV),
2253: (int (*)(const void *,const void *))ndv_compare);
2254: }
1.69 noro 2255: nd_pslen = 2*nd_psn;
1.11 noro 2256: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2257: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2258: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2259: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2260:
1.146 noro 2261: if ( trace && nd_vc )
2262: makesubst(nd_vc,&nd_subst);
2263: else
2264: nd_subst = 0;
2265:
1.11 noro 2266: if ( !nd_red )
1.13 noro 2267: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2268: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2269: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2270: if ( trace ) {
1.69 noro 2271: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2272: ndv_removecont(0,a);
2273: am = nd_ps[i] = ndv_dup(mod,a);
2274: ndv_mod(mod,am);
2275: ndv_removecont(mod,am);
2276: } else {
1.69 noro 2277: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2278: ndv_removecont(mod,a);
1.39 noro 2279: }
1.61 noro 2280: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2281: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2282: nd_psh[i] = r;
1.77 noro 2283: if ( nd_demand ) {
2284: if ( trace ) {
2285: ndv_save(nd_ps_trace[i],i);
2286: nd_ps_trace[i] = 0;
2287: } else {
2288: ndv_save(nd_ps[i],i);
2289: nd_ps[i] = 0;
2290: }
2291: }
1.20 noro 2292: }
2293: }
2294:
1.119 noro 2295: struct order_spec *append_block(struct order_spec *spec,
2296: int nv,int nalg,int ord);
2297:
1.121 noro 2298: extern VECT current_dl_weight_vector_obj;
2299: static VECT prev_weight_vector_obj;
2300:
1.120 noro 2301: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
2302: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
2303: {
1.121 noro 2304: NODE alist,t,s,r0,r,arg;
1.120 noro 2305: VL tv;
2306: P poly;
2307: DP d;
2308: Alg alpha,dp;
2309: DAlg inv,da,hc;
2310: MP m;
1.121 noro 2311: int i,nvar,nalg,n;
1.120 noro 2312: NumberField nf;
2313: LIST f1,f2;
2314: struct order_spec *current_spec;
1.121 noro 2315: VECT obj,obj0;
2316: Obj tmp;
1.120 noro 2317:
2318: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2319: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2320:
2321: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2322: NEXTNODE(alist,t); MKV(tv->v,poly);
2323: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2324: tv->v = tv->v->priv;
2325: }
2326: NEXT(t) = 0;
2327:
2328: /* simplification, makeing polynomials monic */
2329: setfield_dalg(alist);
2330: obj_algtodalg(f,&f1);
2331: for ( t = BDY(f); t; t = NEXT(t) ) {
2332: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2333: hc = (DAlg)BDY(d)->c;
2334: if ( NID(hc) == N_DA ) {
2335: invdalg(hc,&inv);
2336: for ( m = BDY(d); m; m = NEXT(m) ) {
2337: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2338: }
2339: }
2340: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2341: }
2342: obj_dalgtoalg(f1,&f);
2343:
2344: /* append alg vars to the var list */
2345: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2346: NEXT(tv) = av;
2347:
2348: /* append a block to ord */
2349: *ord1p = append_block(ord,nvar,nalg,2);
2350:
2351: /* create generator list */
2352: nf = get_numberfield();
2353: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2354: MKAlg(nf->defpoly[i],dp);
2355: MKNODE(s,dp,t); t = s;
2356: }
2357: MKLIST(f1,t);
2358: *alistp = alist;
2359: algobjtorat(f1,f1p);
1.121 noro 2360:
2361: /* creating a new weight vector */
2362: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2363: n = nvar+nalg+1;
2364: MKVECT(obj,n);
2365: if ( obj0 && obj0->len == nvar )
2366: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2367: else
2368: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2369: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2370: BDY(obj)[n-1] = (pointer)ONE;
2371: arg = mknode(1,obj);
2372: Pdp_set_weight(arg,&tmp);
2373: }
2374:
2375: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2376: {
2377: NODE s,t,u0,u;
2378: P p;
2379: VL tv;
2380: Obj obj,tmp;
2381: NODE arg;
2382:
2383: u0 = 0;
2384: for ( t = r; t; t = NEXT(t) ) {
2385: p = (P)BDY(t);
2386: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2387: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2388: }
2389: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2390: NEXTNODE(u0,u);
2391: BDY(u) = (pointer)p;
2392: }
2393: }
2394: arg = mknode(1,prev_weight_vector_obj);
2395: Pdp_set_weight(arg,&tmp);
2396:
2397: return u0;
1.120 noro 2398: }
2399:
1.63 noro 2400: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2401: {
1.119 noro 2402: VL tv,fv,vv,vc,av;
2403: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2404: int e,max,nvar,i;
1.69 noro 2405: NDV b;
1.119 noro 2406: int ishomo,nalg;
2407: Alg alpha,dp;
1.144 noro 2408: P p,zp;
2409: Q dmy;
1.119 noro 2410: LIST f1,f2;
2411: Obj obj;
2412: NumberField nf;
1.120 noro 2413: struct order_spec *ord1;
1.1 noro 2414:
1.78 noro 2415: if ( !m && Demand ) nd_demand = 1;
2416: else nd_demand = 0;
2417:
1.149 ! noro 2418: if ( DP_Multiple )
! 2419: nd_scale = ((double)DP_Multiple)/(double)(Denominator?Denominator:1);
1.103 noro 2420: #if 0
1.74 noro 2421: ndv_alloc = 0;
1.103 noro 2422: #endif
1.146 noro 2423: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.74 noro 2424: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2425: switch ( ord->id ) {
2426: case 1:
2427: if ( ord->nv != nvar )
2428: error("nd_{gr,f4} : invalid order specification");
2429: break;
2430: default:
2431: break;
2432: }
1.119 noro 2433: nd_nalg = 0;
1.123 noro 2434: av = 0;
1.119 noro 2435: if ( !m ) {
2436: get_algtree((Obj)f,&av);
2437: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2438: nd_ntrans = nvar;
1.119 noro 2439: nd_nalg = nalg;
2440: /* #i -> t#i */
2441: if ( nalg ) {
1.120 noro 2442: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2443: ord = ord1;
2444: f = f1;
1.119 noro 2445: }
1.120 noro 2446: nvar += nalg;
1.119 noro 2447: }
1.32 noro 2448: nd_init_ord(ord);
1.61 noro 2449: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2450: for ( tv = vv; tv; tv = NEXT(tv) ) {
2451: e = getdeg(tv->v,(P)BDY(t));
2452: max = MAX(e,max);
2453: }
2454: nd_setup_parameters(nvar,max);
1.77 noro 2455: ishomo = 1;
1.1 noro 2456: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2457: ptozp((P)BDY(t),1,&dmy,&zp);
2458: b = (pointer)ptondv(CO,vv,zp);
1.77 noro 2459: if ( ishomo )
2460: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2461: if ( m ) ndv_mod(m,b);
1.61 noro 2462: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2463: }
2464: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2465: ndv_setup(m,0,fd0,0);
1.77 noro 2466: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2467: nd_demand = 0;
1.61 noro 2468: x = ndv_reducebase(x);
2469: x = ndv_reduceall(m,x);
1.23 noro 2470: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2471: NEXTNODE(r0,r);
1.61 noro 2472: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2473: }
2474: if ( r0 ) NEXT(r) = 0;
1.121 noro 2475: if ( nalg )
2476: r0 = postprocess_algcoef(av,alist,r0);
1.20 noro 2477: MKLIST(*rp,r0);
1.103 noro 2478: #if 0
1.74 noro 2479: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2480: #endif
1.127 noro 2481: }
2482:
2483: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2484: {
2485: VL tv,fv,vv,vc,av;
2486: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2487: int e,max,nvar,i;
2488: NDV b;
2489: int ishomo,nalg;
2490: Alg alpha,dp;
1.144 noro 2491: P p,zp;
2492: Q dmy;
1.127 noro 2493: LIST f1,f2;
2494: Obj obj;
2495: NumberField nf;
2496: struct order_spec *ord1;
2497:
1.146 noro 2498: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.127 noro 2499: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2500: switch ( ord->id ) {
2501: case 1:
2502: if ( ord->nv != nvar )
2503: error("nd_check : invalid order specification");
2504: break;
2505: default:
2506: break;
2507: }
2508: nd_nalg = 0;
2509: av = 0;
2510: if ( !m ) {
2511: get_algtree((Obj)f,&av);
2512: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2513: nd_ntrans = nvar;
2514: nd_nalg = nalg;
2515: /* #i -> t#i */
2516: if ( nalg ) {
2517: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2518: ord = ord1;
2519: f = f1;
2520: }
2521: nvar += nalg;
2522: }
2523: nd_init_ord(ord);
2524: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2525: for ( tv = vv; tv; tv = NEXT(tv) ) {
2526: e = getdeg(tv->v,(P)BDY(t));
2527: max = MAX(e,max);
2528: }
2529: nd_setup_parameters(nvar,max);
2530: ishomo = 1;
2531: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2532: ptozp((P)BDY(t),1,&dmy,&zp);
2533: b = (pointer)ptondv(CO,vv,zp);
1.127 noro 2534: if ( ishomo )
2535: ishomo = ishomo && ndv_ishomo(b);
2536: if ( m ) ndv_mod(m,b);
2537: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2538: }
2539: if ( fd0 ) NEXT(fd) = 0;
1.128 noro 2540: ndv_setup(m,0,fd0,0);
1.127 noro 2541: for ( x = 0, i = 0; i < nd_psn; i++ )
2542: x = update_base(x,i);
2543: if ( do_check ) {
2544: x = nd_gb(m,ishomo,1);
2545: if ( !x ) {
2546: *rp = 0;
2547: return;
2548: }
2549: } else {
2550: for ( t = x; t; t = NEXT(t) )
2551: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2552: }
2553: x = ndv_reducebase(x);
2554: x = ndv_reduceall(m,x);
2555: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2556: NEXTNODE(r0,r);
2557: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2558: }
2559: if ( r0 ) NEXT(r) = 0;
2560: if ( nalg )
2561: r0 = postprocess_algcoef(av,alist,r0);
2562: MKLIST(*rp,r0);
1.20 noro 2563: }
2564:
1.133 noro 2565: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2566: {
1.119 noro 2567: VL tv,fv,vv,vc,av;
2568: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
1.61 noro 2569: int m,nocheck,nvar,mindex,e,max;
2570: NDV c;
2571: NMV a;
1.144 noro 2572: P p,zp;
2573: Q dmy;
1.61 noro 2574: EPOS oepos;
1.119 noro 2575: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg;
2576: Alg alpha,dp;
2577: P poly;
2578: LIST f1,f2;
2579: Obj obj;
2580: NumberField nf;
1.120 noro 2581: struct order_spec *ord1;
1.145 noro 2582: struct oEGT eg_check,eg0,eg1;
1.20 noro 2583:
1.149 ! noro 2584: if ( DP_Multiple )
! 2585: nd_scale = ((double)DP_Multiple)/(double)(Denominator?Denominator:1);
! 2586:
1.146 noro 2587: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
1.74 noro 2588: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2589: switch ( ord->id ) {
2590: case 1:
2591: if ( ord->nv != nvar )
2592: error("nd_gr_trace : invalid order specification");
2593: break;
2594: default:
2595: break;
2596: }
1.119 noro 2597:
2598: get_algtree((Obj)f,&av);
2599: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
1.121 noro 2600: nd_ntrans = nvar;
1.119 noro 2601: nd_nalg = nalg;
2602: /* #i -> t#i */
2603: if ( nalg ) {
1.120 noro 2604: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2605: ord = ord1;
2606: f = f1;
1.119 noro 2607: }
1.120 noro 2608: nvar += nalg;
1.119 noro 2609:
1.52 noro 2610: nocheck = 0;
2611: mindex = 0;
2612:
1.78 noro 2613: if ( Demand ) nd_demand = 1;
2614: else nd_demand = 0;
2615:
1.52 noro 2616: /* setup modulus */
2617: if ( trace < 0 ) {
2618: trace = -trace;
2619: nocheck = 1;
2620: }
2621: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2622: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2623: for ( tv = vv; tv; tv = NEXT(tv) ) {
2624: e = getdeg(tv->v,(P)BDY(t));
2625: max = MAX(e,max);
1.23 noro 2626: }
1.61 noro 2627: nd_init_ord(ord);
2628: nd_setup_parameters(nvar,max);
2629: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2630: ishomo = 1;
1.61 noro 2631: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.144 noro 2632: ptozp((P)BDY(t),1,&dmy,&zp);
2633: c = (pointer)ptondv(CO,vv,zp);
1.82 noro 2634: if ( ishomo )
2635: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2636: if ( c ) {
2637: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2638: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2639: }
1.61 noro 2640: }
2641: if ( in0 ) NEXT(in) = 0;
2642: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2643: if ( !ishomo && homo ) {
1.122 noro 2644: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
1.61 noro 2645: c = (NDV)BDY(t); len = LEN(c);
2646: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2647: wmax = MAX(TD(DL(a)),wmax);
2648: }
2649: homogenize_order(ord,nvar,&ord1);
1.89 noro 2650: nd_init_ord(ord1);
1.61 noro 2651: nd_setup_parameters(nvar+1,wmax);
2652: for ( t = fd0; t; t = NEXT(t) )
2653: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2654: }
1.52 noro 2655: while ( 1 ) {
1.77 noro 2656: if ( Demand )
2657: nd_demand = 1;
1.128 noro 2658: ndv_setup(m,1,fd0,0);
1.133 noro 2659: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
1.52 noro 2660: if ( !cand ) {
2661: /* failure */
1.61 noro 2662: if ( trace > 1 ) { *rp = 0; return; }
2663: else m = get_lprime(++mindex);
1.52 noro 2664: continue;
2665: }
1.82 noro 2666: if ( !ishomo && homo ) {
1.27 noro 2667: /* dehomogenization */
1.61 noro 2668: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2669: nd_init_ord(ord);
1.61 noro 2670: nd_setup_parameters(nvar,0);
1.27 noro 2671: }
1.77 noro 2672: nd_demand = 0;
1.61 noro 2673: cand = ndv_reducebase(cand);
2674: cand = ndv_reduceall(0,cand);
1.78 noro 2675: cbpe = nd_bpe;
1.61 noro 2676: if ( nocheck )
2677: break;
1.145 noro 2678: get_eg(&eg0);
1.61 noro 2679: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2680: /* success */
2681: break;
2682: else if ( trace > 1 ) {
2683: /* failure */
2684: *rp = 0; return;
1.61 noro 2685: } else {
1.52 noro 2686: /* try the next modulus */
2687: m = get_lprime(++mindex);
1.61 noro 2688: /* reset the parameters */
1.82 noro 2689: if ( !ishomo && homo ) {
1.89 noro 2690: nd_init_ord(ord1);
1.61 noro 2691: nd_setup_parameters(nvar+1,wmax);
2692: } else {
2693: nd_init_ord(ord);
2694: nd_setup_parameters(nvar,max);
2695: }
2696: }
1.52 noro 2697: }
1.145 noro 2698: get_eg(&eg1); init_eg(&eg_check); add_eg(&eg_check,&eg0,&eg1);
2699: if ( DP_Print )
2700: fprintf(asir_out,"check=%fsec\n",eg_check.exectime+eg_check.gctime);
1.27 noro 2701: /* dp->p */
1.61 noro 2702: nd_bpe = cbpe;
1.73 noro 2703: nd_setup_parameters(nd_nvar,0);
1.121 noro 2704: for ( r = cand; r; r = NEXT(r) )
1.119 noro 2705: BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.121 noro 2706: if ( nalg )
2707: cand = postprocess_algcoef(av,alist,cand);
1.27 noro 2708: MKLIST(*rp,cand);
1.1 noro 2709: }
2710:
1.61 noro 2711: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2712: {
1.61 noro 2713: UINT *d;
1.57 noro 2714: int i,j,l,s,ord_l;
1.43 noro 2715: struct order_pair *op;
1.1 noro 2716:
2717: d = dl->d;
1.41 noro 2718: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2719: if ( nd_blockmask ) {
2720: l = nd_blockmask->n;
2721: op = nd_blockmask->order_pair;
2722: for ( j = 0, s = 0; j < l; j++ ) {
2723: ord_l = op[j].length;
1.57 noro 2724: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2725: }
2726: TD(r) = ndl_weight(r);
1.61 noro 2727: ndl_weight_mask(r);
1.43 noro 2728: } else {
1.56 noro 2729: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2730: TD(r) = ndl_weight(r);
2731: }
1.1 noro 2732: }
2733:
1.61 noro 2734: DL ndltodl(int n,UINT *ndl)
1.1 noro 2735: {
2736: DL dl;
2737: int *d;
1.57 noro 2738: int i,j,l,s,ord_l;
1.43 noro 2739: struct order_pair *op;
1.1 noro 2740:
2741: NEWDL(dl,n);
1.34 noro 2742: dl->td = TD(ndl);
1.1 noro 2743: d = dl->d;
1.43 noro 2744: if ( nd_blockmask ) {
2745: l = nd_blockmask->n;
2746: op = nd_blockmask->order_pair;
2747: for ( j = 0, s = 0; j < l; j++ ) {
2748: ord_l = op[j].length;
1.57 noro 2749: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2750: }
2751: } else {
1.56 noro 2752: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2753: }
1.1 noro 2754: return dl;
2755: }
2756:
1.61 noro 2757: void ndl_print(UINT *dl)
1.1 noro 2758: {
2759: int n;
1.57 noro 2760: int i,j,l,ord_l,s,s0;
1.43 noro 2761: struct order_pair *op;
1.1 noro 2762:
2763: n = nd_nvar;
2764: printf("<<");
1.43 noro 2765: if ( nd_blockmask ) {
2766: l = nd_blockmask->n;
2767: op = nd_blockmask->order_pair;
2768: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2769: ord_l = op[j].length;
1.57 noro 2770: for ( i = 0; i < ord_l; i++, s++ )
2771: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2772: }
2773: } else {
1.56 noro 2774: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2775: }
1.1 noro 2776: printf(">>");
2777: }
2778:
2779: void nd_print(ND p)
2780: {
2781: NM m;
2782:
2783: if ( !p )
2784: printf("0\n");
2785: else {
2786: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2787: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2788: else printf("+%d*",CM(m));
1.14 noro 2789: ndl_print(DL(m));
1.1 noro 2790: }
2791: printf("\n");
2792: }
2793: }
2794:
1.113 noro 2795: void nd_print_q(ND p)
1.16 noro 2796: {
2797: NM m;
2798:
2799: if ( !p )
2800: printf("0\n");
2801: else {
2802: for ( m = BDY(p); m; m = NEXT(m) ) {
2803: printf("+");
1.113 noro 2804: printexpr(CO,(Obj)CQ(m));
1.16 noro 2805: printf("*");
2806: ndl_print(DL(m));
2807: }
2808: printf("\n");
2809: }
2810: }
2811:
1.1 noro 2812: void ndp_print(ND_pairs d)
2813: {
2814: ND_pairs t;
2815:
1.34 noro 2816: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2817: printf("\n");
2818: }
2819:
1.20 noro 2820: void nd_removecont(int mod,ND p)
1.16 noro 2821: {
2822: int i,n;
1.113 noro 2823: Q *w;
2824: Q dvr,t;
1.16 noro 2825: NM m;
1.21 noro 2826: struct oVECT v;
1.113 noro 2827: N q,r;
1.16 noro 2828:
1.71 noro 2829: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2830: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2831: else {
2832: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
1.113 noro 2833: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2834: v.len = n;
2835: v.body = (pointer *)w;
1.113 noro 2836: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.146 noro 2837: removecont_array((P *)w,n);
1.113 noro 2838: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2839: }
2840: }
2841:
1.21 noro 2842: void nd_removecont2(ND p1,ND p2)
2843: {
2844: int i,n1,n2,n;
1.113 noro 2845: Q *w;
2846: Q dvr,t;
1.21 noro 2847: NM m;
2848: struct oVECT v;
1.113 noro 2849: N q,r;
1.21 noro 2850:
2851: if ( !p1 ) {
2852: nd_removecont(0,p2); return;
2853: } else if ( !p2 ) {
2854: nd_removecont(0,p1); return;
2855: }
2856: n1 = nd_length(p1);
2857: n2 = nd_length(p2);
2858: n = n1+n2;
1.113 noro 2859: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2860: v.len = n;
2861: v.body = (pointer *)w;
1.113 noro 2862: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2863: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.146 noro 2864: removecont_array((P *)w,n);
1.113 noro 2865: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2866: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 2867: }
2868:
1.20 noro 2869: void ndv_removecont(int mod,NDV p)
1.16 noro 2870: {
1.146 noro 2871: int i,len,all_p;
2872: Q *c;
2873: P *w;
1.113 noro 2874: Q dvr,t;
1.146 noro 2875: P g,cont,tp;
1.16 noro 2876: NMV m;
2877:
1.71 noro 2878: if ( mod == -1 )
2879: ndv_mul_c(mod,p,_invsf(HCM(p)));
2880: else if ( mod )
1.20 noro 2881: ndv_mul_c(mod,p,invm(HCM(p),mod));
2882: else {
2883: len = p->len;
1.146 noro 2884: w = (P *)ALLOCA(len*sizeof(P));
2885: c = (Q *)ALLOCA(len*sizeof(Q));
2886: for ( m = BDY(p), all_p = 1, i = 0; i < len; NMV_ADV(m), i++ ) {
2887: ptozp(CP(m),1,&c[i],&w[i]);
2888: all_p = all_p && !NUM(w[i]);
2889: }
2890: if ( all_p ) {
2891: qltozl(c,len,&dvr); heu_nezgcdnpz(nd_vc,w,len,&g);
2892: mulp(nd_vc,(P)dvr,g,&cont);
2893: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2894: divsp(nd_vc,CP(m),cont,&tp); CP(m) = tp;
2895: }
2896: } else {
2897: sortbynm((Q *)c,len);
2898: qltozl((Q *)c,len,&dvr);
2899: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2900: divsp(nd_vc,CP(m),(P)dvr,&tp); CP(m) = tp;
2901: }
1.113 noro 2902: }
1.16 noro 2903: }
1.21 noro 2904: }
2905:
1.61 noro 2906: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2907: {
2908: int len,i,max;
2909: NMV m,mr0,mr,t;
2910:
2911: len = p->len;
2912: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2913: max = MAX(max,TD(DL(m)));
2914: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2915: m = (NMV)((char *)mr0+(len-1)*oadv);
2916: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2917: t = (NMV)ALLOCA(nmv_adv);
2918: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2919: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
1.113 noro 2920: CQ(mr) = CQ(m);
1.61 noro 2921: ndl_copy(DL(t),DL(mr));
2922: }
2923: NV(p)++;
2924: BDY(p) = mr0;
2925: }
2926:
1.45 noro 2927: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2928: {
1.45 noro 2929: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.113 noro 2930: Q *w;
2931: Q dvr,t;
1.23 noro 2932: NMV m,r;
2933:
2934: len = p->len;
2935: newnvar = nd_nvar-1;
1.48 noro 2936: newexporigin = nd_get_exporigin(ord);
1.45 noro 2937: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2938: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2939: ndl_dehomogenize(DL(m));
1.23 noro 2940: if ( newwpd != nd_wpd ) {
1.90 noro 2941: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2942: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.113 noro 2943: CQ(r) = CQ(m);
1.45 noro 2944: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2945: adj = nd_exporigin-newexporigin;
2946: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2947: }
2948: }
2949: NV(p)--;
2950: }
2951:
1.146 noro 2952: void removecont_array(P *p,int n)
2953: {
2954: int all_p,all_q,i;
2955: Q *c;
2956: P *w;
2957: P t,s;
2958:
2959: for ( all_q = 1, i = 0; i < n; i++ )
2960: all_q = all_q && NUM(p[i]);
2961: if ( all_q ) {
2962: removecont_array_q((Q *)p,n);
2963: } else {
2964: c = (Q *)ALLOCA(n*sizeof(Q));
2965: w = (P *)ALLOCA(n*sizeof(P));
2966: for ( i = 0; i < n; i++ ) {
2967: ptozp(p[i],1,&c[i],&w[i]);
2968: }
2969: removecont_array_q(c,n);
2970: heu_nezgcdnpz(nd_vc,w,n,&t);
2971: for ( i = 0; i < n; i++ ) {
2972: divsp(nd_vc,w[i],t,&s); mulp(nd_vc,s,(P)c[i],&p[i]);
2973: }
2974: }
2975: }
2976:
2977: void removecont_array_q(Q *c,int n)
1.21 noro 2978: {
2979: struct oVECT v;
1.113 noro 2980: Q d0,d1,a,u,u1,gcd;
1.110 noro 2981: int i,j;
1.113 noro 2982: N qn,rn,gn;
2983: Q *q,*r;
1.21 noro 2984:
1.113 noro 2985: q = (Q *)ALLOCA(n*sizeof(Q));
2986: r = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2987: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
1.113 noro 2988: igcdv_estimate(&v,&d0);
2989: for ( i = 0; i < n; i++ ) {
2990: divn(NM(c[i]),NM(d0),&qn,&rn);
2991: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2992: NTOQ(rn,SGN(c[i]),r[i]);
2993: }
1.34 noro 2994: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2995: if ( i < n ) {
2996: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
1.113 noro 2997: igcdv(&v,&d1);
2998: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2999: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
1.21 noro 3000: for ( i = 0; i < n; i++ ) {
1.113 noro 3001: mulq(a,q[i],&u);
3002: if ( r[i] ) {
3003: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
3004: addq(u,u1,&q[i]);
3005: } else
3006: q[i] = u;
1.21 noro 3007: }
3008: }
1.34 noro 3009: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 3010: }
3011:
1.19 noro 3012: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 3013: {
3014: NM m;
3015: int c,c1;
3016:
1.34 noro 3017: if ( !p ) return;
1.115 noro 3018: if ( mul == 1 ) return;
1.71 noro 3019: if ( mod == -1 )
3020: for ( m = BDY(p); m; m = NEXT(m) )
3021: CM(m) = _mulsf(CM(m),mul);
3022: else
3023: for ( m = BDY(p); m; m = NEXT(m) ) {
3024: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3025: }
1.1 noro 3026: }
3027:
1.146 noro 3028: void nd_mul_c_q(ND p,P mul)
1.16 noro 3029: {
3030: NM m;
1.146 noro 3031: P c;
1.16 noro 3032:
1.113 noro 3033: if ( !p ) return;
1.115 noro 3034: if ( UNIQ(mul) ) return;
1.16 noro 3035: for ( m = BDY(p); m; m = NEXT(m) ) {
1.146 noro 3036: mulp(nd_vc,CP(m),mul,&c); CP(m) = c;
1.16 noro 3037: }
3038: }
3039:
1.61 noro 3040: void nd_mul_c_p(VL vl,ND p,P mul)
3041: {
3042: NM m;
3043: P c;
3044:
3045: if ( !p ) return;
3046: for ( m = BDY(p); m; m = NEXT(m) ) {
3047: mulp(vl,CP(m),mul,&c); CP(m) = c;
3048: }
3049: }
3050:
1.1 noro 3051: void nd_free(ND p)
3052: {
3053: NM t,s;
3054:
1.34 noro 3055: if ( !p ) return;
1.1 noro 3056: t = BDY(p);
3057: while ( t ) {
3058: s = NEXT(t);
3059: FREENM(t);
3060: t = s;
3061: }
3062: FREEND(p);
3063: }
3064:
1.23 noro 3065: void ndv_free(NDV p)
3066: {
3067: GC_free(BDY(p));
3068: }
3069:
1.61 noro 3070: void nd_append_red(UINT *d,int i)
1.1 noro 3071: {
1.13 noro 3072: RHist m,m0;
1.1 noro 3073: int h;
3074:
1.13 noro 3075: NEWRHist(m);
1.34 noro 3076: h = ndl_hash_value(d);
1.13 noro 3077: m->index = i;
1.14 noro 3078: ndl_copy(d,DL(m));
1.1 noro 3079: NEXT(m) = nd_red[h];
3080: nd_red[h] = m;
3081: }
3082:
1.61 noro 3083: UINT *ndv_compute_bound(NDV p)
1.1 noro 3084: {
1.61 noro 3085: UINT *d1,*d2,*t;
3086: UINT u;
1.57 noro 3087: int i,j,k,l,len,ind;
1.45 noro 3088: NMV m;
1.1 noro 3089:
3090: if ( !p )
3091: return 0;
1.61 noro 3092: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3093: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 3094: len = LEN(p);
3095: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3096: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3097: ndl_lcm(DL(m),d1,d2);
1.1 noro 3098: t = d1; d1 = d2; d2 = t;
3099: }
1.12 noro 3100: l = nd_nvar+31;
1.61 noro 3101: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 3102: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3103: u = d1[i];
3104: k = (nd_epw-1)*nd_bpe;
3105: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3106: t[ind] = (u>>k)&nd_mask0;
3107: }
3108: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 3109: return t;
3110: }
3111:
1.99 noro 3112: UINT *nd_compute_bound(ND p)
3113: {
3114: UINT *d1,*d2,*t;
3115: UINT u;
3116: int i,j,k,l,len,ind;
3117: NM m;
3118:
3119: if ( !p )
3120: return 0;
3121: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3122: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3123: len = LEN(p);
3124: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3125: for ( m = NEXT(m); m; m = NEXT(m) ) {
3126: ndl_lcm(DL(m),d1,d2);
3127: t = d1; d1 = d2; d2 = t;
3128: }
3129: l = nd_nvar+31;
3130: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3131: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3132: u = d1[i];
3133: k = (nd_epw-1)*nd_bpe;
3134: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3135: t[ind] = (u>>k)&nd_mask0;
3136: }
3137: for ( ; ind < l; ind++ ) t[ind] = 0;
3138: return t;
3139: }
3140:
1.48 noro 3141: int nd_get_exporigin(struct order_spec *ord)
3142: {
1.51 noro 3143: switch ( ord->id ) {
1.96 noro 3144: case 0: case 2:
1.48 noro 3145: return 1;
1.41 noro 3146: case 1:
3147: /* block order */
1.43 noro 3148: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 3149: return ord->ord.block.length+1;
1.96 noro 3150: case 3:
3151: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 3152: }
1.48 noro 3153: }
3154:
1.61 noro 3155: void nd_setup_parameters(int nvar,int max) {
1.79 noro 3156: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 3157: struct order_pair *op;
1.48 noro 3158:
1.73 noro 3159: nd_nvar = nvar;
1.79 noro 3160: if ( max ) {
1.83 noro 3161: /* XXX */
3162: if ( do_weyl ) nd_bpe = 32;
3163: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 3164: else if ( max < 4 ) nd_bpe = 2;
3165: else if ( max < 8 ) nd_bpe = 3;
3166: else if ( max < 16 ) nd_bpe = 4;
3167: else if ( max < 32 ) nd_bpe = 5;
3168: else if ( max < 64 ) nd_bpe = 6;
3169: else if ( max < 256 ) nd_bpe = 8;
3170: else if ( max < 1024 ) nd_bpe = 10;
3171: else if ( max < 65536 ) nd_bpe = 16;
3172: else nd_bpe = 32;
3173: }
1.61 noro 3174: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 3175: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3176: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 3177: wpd = nd_exporigin+elen;
3178: if ( wpd != nd_wpd ) {
3179: nd_free_private_storage();
3180: nd_wpd = wpd;
3181: }
1.1 noro 3182: if ( nd_bpe < 32 ) {
3183: nd_mask0 = (1<<nd_bpe)-1;
3184: } else {
3185: nd_mask0 = 0xffffffff;
3186: }
3187: bzero(nd_mask,sizeof(nd_mask));
3188: nd_mask1 = 0;
3189: for ( i = 0; i < nd_epw; i++ ) {
3190: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3191: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3192: }
1.90 noro 3193: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 3194: nd_epos = nd_create_epos(nd_ord);
1.43 noro 3195: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 noro 3196: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3197: }
3198:
1.103 noro 3199: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3200: {
1.37 noro 3201: int i,obpe,oadv,h;
1.71 noro 3202: static NM prev_nm_free_list;
3203: static ND_pairs prev_ndp_free_list;
1.13 noro 3204: RHist mr0,mr;
3205: RHist r;
1.37 noro 3206: RHist *old_red;
1.71 noro 3207: ND_pairs s0,s,t;
1.43 noro 3208: EPOS oepos;
1.15 noro 3209:
1.1 noro 3210: obpe = nd_bpe;
1.11 noro 3211: oadv = nmv_adv;
1.43 noro 3212: oepos = nd_epos;
1.63 noro 3213: if ( obpe < 2 ) nd_bpe = 2;
3214: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 3215: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 3216: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 3217: else if ( obpe < 6 ) nd_bpe = 6;
3218: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 3219: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 3220: else if ( obpe < 16 ) nd_bpe = 16;
3221: else if ( obpe < 32 ) nd_bpe = 32;
3222: else error("nd_reconstruct : exponent too large");
1.5 noro 3223:
1.73 noro 3224: nd_setup_parameters(nd_nvar,0);
1.1 noro 3225: prev_nm_free_list = _nm_free_list;
3226: prev_ndp_free_list = _ndp_free_list;
3227: _nm_free_list = 0;
3228: _ndp_free_list = 0;
1.53 noro 3229: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3230: if ( trace )
3231: for ( i = nd_psn-1; i >= 0; i-- )
3232: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 3233: s0 = 0;
3234: for ( t = d; t; t = NEXT(t) ) {
3235: NEXTND_pairs(s0,s);
3236: s->i1 = t->i1;
3237: s->i2 = t->i2;
1.14 noro 3238: SG(s) = SG(t);
1.61 noro 3239: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 3240: }
1.37 noro 3241:
3242: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 3243: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 3244: old_red[i] = nd_red[i];
3245: nd_red[i] = 0;
3246: }
3247: for ( i = 0; i < REDTAB_LEN; i++ )
3248: for ( r = old_red[i]; r; r = NEXT(r) ) {
3249: NEWRHist(mr);
1.13 noro 3250: mr->index = r->index;
1.20 noro 3251: SG(mr) = SG(r);
1.61 noro 3252: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 3253: h = ndl_hash_value(DL(mr));
3254: NEXT(mr) = nd_red[h];
3255: nd_red[h] = mr;
1.6 noro 3256: }
1.37 noro 3257: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3258: old_red = 0;
1.11 noro 3259: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 3260: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3261: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3262: nd_psh[i] = r;
1.11 noro 3263: }
1.1 noro 3264: if ( s0 ) NEXT(s) = 0;
3265: prev_nm_free_list = 0;
3266: prev_ndp_free_list = 0;
1.71 noro 3267: #if 0
1.1 noro 3268: GC_gcollect();
1.71 noro 3269: #endif
1.1 noro 3270: return s0;
3271: }
3272:
1.61 noro 3273: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3274: {
1.57 noro 3275: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3276: struct order_pair *op;
1.1 noro 3277:
3278: n = nd_nvar;
1.61 noro 3279: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3280: omask0 = (1<<obpe)-1;
1.34 noro 3281: TD(r) = TD(d);
1.41 noro 3282: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3283: if ( nd_blockmask ) {
3284: l = nd_blockmask->n;
3285: op = nd_blockmask->order_pair;
3286: for ( i = 1; i < nd_exporigin; i++ )
3287: r[i] = d[i];
3288: for ( j = 0, s = 0; j < l; j++ ) {
3289: ord_l = op[j].length;
1.57 noro 3290: for ( i = 0; i < ord_l; i++, s++ ) {
3291: ei = GET_EXP_OLD(d,s);
3292: PUT_EXP(r,s,ei);
3293: }
1.1 noro 3294: }
1.43 noro 3295: } else {
1.56 noro 3296: for ( i = 0; i < n; i++ ) {
3297: ei = GET_EXP_OLD(d,i);
3298: PUT_EXP(r,i,ei);
3299: }
1.1 noro 3300: }
3301: }
1.3 noro 3302:
1.6 noro 3303: ND nd_copy(ND p)
3304: {
3305: NM m,mr,mr0;
1.41 noro 3306: int c,n;
1.6 noro 3307: ND r;
3308:
3309: if ( !p )
3310: return 0;
3311: else {
3312: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3313: NEXTNM(mr0,mr);
1.14 noro 3314: CM(mr) = CM(m);
3315: ndl_copy(DL(m),DL(mr));
1.6 noro 3316: }
3317: NEXT(mr) = 0;
1.31 noro 3318: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3319: SG(r) = SG(p);
1.6 noro 3320: return r;
3321: }
3322: }
3323:
1.53 noro 3324: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3325: {
1.148 noro 3326: NM m1,m2;
1.11 noro 3327: NDV p1,p2;
3328: ND t1,t2;
1.61 noro 3329: UINT *lcm;
1.148 noro 3330: P gp,tp;
3331: Q g,t;
1.31 noro 3332: int td;
1.11 noro 3333:
1.77 noro 3334: if ( !mod && nd_demand ) {
3335: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3336: } else {
1.77 noro 3337: if ( trace ) {
3338: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3339: } else {
3340: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3341: }
1.20 noro 3342: }
1.34 noro 3343: lcm = LCM(p);
1.148 noro 3344: NEWNM(m1); ndl_sub(lcm,HDL(p1),DL(m1));
3345: if ( ndl_check_bound2(p->i1,DL(m1)) ) {
3346: FREENM(m1); return 0;
3347: }
3348: NEWNM(m2); ndl_sub(lcm,HDL(p2),DL(m2));
3349: if ( ndl_check_bound2(p->i2,DL(m2)) ) {
3350: FREENM(m1); FREENM(m2); return 0;
3351: }
3352:
3353: if ( mod == -1 ) {
3354: CM(m1) = HCM(p2); CM(m2) = _chsgnsf(HCM(p1));
3355: } else if ( mod ) {
3356: CM(m1) = HCM(p2); CM(m2) = mod-HCM(p1);
3357: } else if ( nd_vc ) {
3358: ezgcdpz(nd_vc,HCP(p1),HCP(p2),&gp);
3359: divsp(nd_vc,HCP(p2),gp,&CP(m1));
3360: divsp(nd_vc,HCP(p1),gp,&tp); chsgnp(tp,&CP(m2));
3361: } else {
3362: igcd_cofactor(HCQ(p1),HCQ(p2),&g,&t,&CQ(m1)); chsgnq(t,&CQ(m2));
1.11 noro 3363: }
1.148 noro 3364: t1 = ndv_mul_nm(mod,m1,p1); t2 = ndv_mul_nm(mod,m2,p2);
1.31 noro 3365: *rp = nd_add(mod,t1,t2);
1.148 noro 3366: FREENM(m1); FREENM(m2);
1.11 noro 3367: return 1;
3368: }
3369:
1.19 noro 3370: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3371: {
3372: NMV m;
3373: int c,c1,len,i;
3374:
1.34 noro 3375: if ( !p ) return;
1.14 noro 3376: len = LEN(p);
1.71 noro 3377: if ( mod == -1 )
3378: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3379: CM(m) = _mulsf(CM(m),mul);
3380: else
3381: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3382: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3383: }
1.11 noro 3384: }
3385:
1.113 noro 3386: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3387: {
3388: NMV m;
1.113 noro 3389: Q c;
1.16 noro 3390: int len,i;
3391:
1.34 noro 3392: if ( !p ) return;
1.16 noro 3393: len = LEN(p);
3394: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 3395: mulq(CQ(m),mul,&c); CQ(m) = c;
1.16 noro 3396: }
3397: }
3398:
1.55 noro 3399: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3400: int n2,i,j,l,n,tlen;
1.61 noro 3401: UINT *d0;
1.55 noro 3402: NM *tab,*psum;
3403: ND s,r;
3404: NM t;
3405: NMV m1;
3406:
3407: if ( !p ) return 0;
3408: n = NV(p); n2 = n>>1;
3409: d0 = DL(m0);
3410: l = LEN(p);
3411: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3412: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3413: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3414: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3415: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3416: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3417: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3418: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3419: for ( j = 0; j < tlen; j++ ) {
3420: if ( tab[j] ) {
3421: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3422: }
3423: }
3424: }
3425: for ( i = tlen-1, r = 0; i >= 0; i-- )
3426: if ( psum[i] ) {
3427: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3428: MKND(n,psum[i],j,s);
3429: r = nd_add(mod,r,s);
3430: }
1.56 noro 3431: if ( r ) SG(r) = SG(p)+TD(d0);
3432: return r;
1.55 noro 3433: }
3434:
1.56 noro 3435: /* product of monomials */
3436: /* XXX block order is not handled correctly */
3437:
1.55 noro 3438: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3439: {
1.56 noro 3440: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3441: UINT *d0,*d1,*d,*dt,*ctab;
1.113 noro 3442: Q *ctab_q;
3443: Q q,q1;
1.61 noro 3444: UINT c0,c1,c;
1.55 noro 3445: NM *p;
3446: NM m,t;
3447:
3448: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3449: if ( !m0 || !m1 ) return;
3450: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3451: NEWNM(m); d = DL(m);
1.56 noro 3452: if ( mod ) {
3453: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3454: } else
1.113 noro 3455: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3456: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3457: homo = n&1 ? 1 : 0;
3458: if ( homo ) {
3459: /* offset of h-degree */
3460: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3461: PUT_EXP(DL(m),n-1,h);
3462: TD(DL(m)) = h;
1.61 noro 3463: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3464: }
3465: tab[0] = m;
3466: NEWNM(m); d = DL(m);
1.57 noro 3467: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3468: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3469: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3470: /* xi^a*(Di^k*xi^l)*Di^b */
3471: a += l; b += k;
1.56 noro 3472: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3473: if ( !k || !l ) {
3474: for ( j = 0; j < curlen; j++ )
1.56 noro 3475: if ( t = tab[j] ) {
3476: dt = DL(t);
3477: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3478: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3479: }
3480: curlen *= k+1;
3481: continue;
3482: }
3483: min = MIN(k,l);
1.56 noro 3484: if ( mod ) {
1.61 noro 3485: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3486: mkwcm(k,l,mod,ctab);
3487: } else {
1.113 noro 3488: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3489: mkwc(k,l,ctab_q);
1.56 noro 3490: }
1.57 noro 3491: for ( j = min; j >= 0; j-- ) {
1.56 noro 3492: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3493: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3494: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3495: if ( homo ) {
3496: TD(d) = s;
1.56 noro 3497: PUT_EXP(d,n-1,s-h);
1.55 noro 3498: } else TD(d) = h;
1.61 noro 3499: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3500: if ( mod ) c = ctab[j];
1.113 noro 3501: else q = ctab_q[j];
1.57 noro 3502: p = tab+curlen*j;
3503: if ( j == 0 ) {
3504: for ( u = 0; u < curlen; u++, p++ ) {
3505: if ( tab[u] ) {
3506: ndl_addto(DL(tab[u]),d);
3507: if ( mod ) {
3508: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3509: } else {
1.113 noro 3510: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
1.57 noro 3511: }
3512: }
1.56 noro 3513: }
1.57 noro 3514: } else {
3515: for ( u = 0; u < curlen; u++, p++ ) {
3516: if ( tab[u] ) {
3517: NEWNM(t);
3518: ndl_add(DL(tab[u]),d,DL(t));
3519: if ( mod ) {
3520: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3521: } else
1.113 noro 3522: mulq(CQ(tab[u]),q,&CQ(t));
1.57 noro 3523: *p = t;
3524: }
1.55 noro 3525: }
3526: }
3527: }
3528: curlen *= k+1;
3529: }
3530: FREENM(m);
3531: }
3532:
1.63 noro 3533: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3534: {
3535: NM mr,mr0;
3536: NMV m;
3537: UINT *d,*dt,*dm;
3538: int c,n,td,i,c1,c2,len;
1.113 noro 3539: Q q;
1.63 noro 3540: ND r;
3541:
3542: if ( !p ) return 0;
3543: else {
3544: n = NV(p); m = BDY(p);
3545: d = DL(m0);
3546: len = LEN(p);
3547: mr0 = 0;
3548: td = TD(d);
3549: c = CM(m0);
3550: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3551: NEXTNM(mr0,mr);
3552: CM(mr) = 1;
3553: ndl_add(DL(m),d,DL(mr));
3554: }
3555: NEXT(mr) = 0;
3556: MKND(NV(p),mr0,len,r);
3557: SG(r) = SG(p) + TD(d);
3558: return r;
3559: }
3560: }
3561:
1.55 noro 3562: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3563: {
3564: NM mr,mr0;
3565: NMV m;
1.61 noro 3566: UINT *d,*dt,*dm;
1.9 noro 3567: int c,n,td,i,c1,c2,len;
1.146 noro 3568: P q;
1.9 noro 3569: ND r;
3570:
1.34 noro 3571: if ( !p ) return 0;
1.55 noro 3572: else if ( do_weyl )
1.71 noro 3573: if ( mod == -1 )
3574: error("ndv_mul_nm : not implemented (weyl)");
3575: else
3576: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3577: else {
3578: n = NV(p); m = BDY(p);
1.34 noro 3579: d = DL(m0);
1.14 noro 3580: len = LEN(p);
1.9 noro 3581: mr0 = 0;
1.34 noro 3582: td = TD(d);
1.71 noro 3583: if ( mod == -1 ) {
3584: c = CM(m0);
3585: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3586: NEXTNM(mr0,mr);
3587: CM(mr) = _mulsf(CM(m),c);
3588: ndl_add(DL(m),d,DL(mr));
3589: }
3590: } else if ( mod ) {
1.16 noro 3591: c = CM(m0);
3592: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3593: NEXTNM(mr0,mr);
3594: c1 = CM(m);
1.19 noro 3595: DMAR(c1,c,0,mod,c2);
1.16 noro 3596: CM(mr) = c2;
3597: ndl_add(DL(m),d,DL(mr));
3598: }
3599: } else {
1.146 noro 3600: q = CP(m0);
1.16 noro 3601: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3602: NEXTNM(mr0,mr);
1.146 noro 3603: mulp(nd_vc,CP(m),q,&CP(mr));
1.16 noro 3604: ndl_add(DL(m),d,DL(mr));
3605: }
1.4 noro 3606: }
1.9 noro 3607: NEXT(mr) = 0;
1.31 noro 3608: MKND(NV(p),mr0,len,r);
1.34 noro 3609: SG(r) = SG(p) + TD(d);
1.9 noro 3610: return r;
1.4 noro 3611: }
3612: }
3613:
1.104 noro 3614: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 3615: {
3616: NM mq0,mq;
1.102 noro 3617: NMV tm;
1.113 noro 3618: Q q;
1.104 noro 3619: int i,nv,sg,c,c1,c2,hindex;
3620: ND p,t,r;
3621: N tnm;
3622:
1.124 noro 3623: if ( bucket->m < 0 ) return 0;
1.99 noro 3624: else {
1.104 noro 3625: nv = NV(d);
1.99 noro 3626: mq0 = 0;
1.102 noro 3627: tm = (NMV)ALLOCA(nmv_adv);
1.104 noro 3628: while ( 1 ) {
1.113 noro 3629: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.104 noro 3630: if ( hindex < 0 ) break;
3631: p = bucket->body[hindex];
1.99 noro 3632: NEXTNM(mq0,mq);
3633: ndl_sub(HDL(p),HDL(d),DL(mq));
1.102 noro 3634: ndl_copy(DL(mq),DL(tm));
3635: if ( mod ) {
3636: c1 = invm(HCM(d),mod); c2 = HCM(p);
3637: DMAR(c1,c2,0,mod,c); CM(mq) = c;
3638: CM(tm) = mod-c;
3639: } else {
1.113 noro 3640: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
3641: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
3642: chsgnq(CQ(mq),&CQ(tm));
1.102 noro 3643: }
3644: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
1.104 noro 3645: bucket->body[hindex] = nd_remove_head(p);
3646: t = nd_remove_head(t);
3647: add_pbucket(mod,bucket,t);
3648: }
3649: if ( !mq0 )
3650: r = 0;
3651: else {
3652: NEXT(mq) = 0;
3653: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
3654: MKND(nv,mq0,i,r);
3655: /* XXX */
3656: SG(r) = HTD(r);
1.99 noro 3657: }
3658: return r;
3659: }
3660: }
3661:
1.43 noro 3662: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3663: {
1.13 noro 3664: NMV m,mr,mr0,t;
3665: int len,i,k;
1.11 noro 3666:
1.61 noro 3667: if ( !p ) return;
3668: m = BDY(p); len = LEN(p);
3669: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3670: m = (NMV)((char *)mr0+(len-1)*oadv);
3671: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3672: t = (NMV)ALLOCA(nmv_adv);
3673: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
1.113 noro 3674: CQ(t) = CQ(m);
1.61 noro 3675: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3676: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
1.113 noro 3677: CQ(mr) = CQ(t);
1.61 noro 3678: ndl_copy(DL(t),DL(mr));
3679: }
3680: BDY(p) = mr0;
3681: }
3682:
3683: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3684: {
3685: NMV m,mr,mr0;
3686: int len,i;
3687: NDV r;
1.11 noro 3688:
1.61 noro 3689: if ( !p ) return 0;
3690: m = BDY(p); len = LEN(p);
3691: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3692: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3693: ndl_zero(DL(mr));
3694: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
1.113 noro 3695: CQ(mr) = CQ(m);
1.11 noro 3696: }
1.61 noro 3697: MKNDV(NV(p),mr0,len,r);
3698: SG(r) = SG(p);
3699: return r;
1.11 noro 3700: }
3701:
1.61 noro 3702: /* duplicate p */
3703:
3704: NDV ndv_dup(int mod,NDV p)
1.3 noro 3705: {
3706: NDV d;
1.61 noro 3707: NMV t,m,m0;
1.3 noro 3708: int i,len;
3709:
1.34 noro 3710: if ( !p ) return 0;
1.31 noro 3711: len = LEN(p);
1.34 noro 3712: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3713: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3714: ndl_copy(DL(t),DL(m));
1.113 noro 3715: CQ(m) = CQ(t);
1.3 noro 3716: }
3717: MKNDV(NV(p),m0,len,d);
1.23 noro 3718: SG(d) = SG(p);
3719: return d;
3720: }
3721:
1.63 noro 3722: ND nd_dup(ND p)
3723: {
3724: ND d;
3725: NM t,m,m0;
3726:
3727: if ( !p ) return 0;
3728: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3729: NEXTNM(m0,m);
3730: ndl_copy(DL(t),DL(m));
1.113 noro 3731: CQ(m) = CQ(t);
1.63 noro 3732: }
3733: if ( m0 ) NEXT(m) = 0;
3734: MKND(NV(p),m0,LEN(p),d);
3735: SG(d) = SG(p);
3736: return d;
3737: }
3738:
1.61 noro 3739: /* XXX if p->len == 0 then it represents 0 */
3740:
3741: void ndv_mod(int mod,NDV p)
3742: {
3743: NMV t,d;
1.125 noro 3744: int r,s,u;
1.61 noro 3745: int i,len,dlen;
1.147 noro 3746: P cp;
1.146 noro 3747: Q c;
1.71 noro 3748: Obj gfs;
1.61 noro 3749:
3750: if ( !p ) return;
3751: len = LEN(p);
3752: dlen = 0;
1.71 noro 3753: if ( mod == -1 )
3754: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3755: simp_ff((Obj)CP(t),&gfs);
3756: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3757: CM(d) = r;
3758: ndl_copy(DL(t),DL(d));
3759: NMV_ADV(d);
3760: dlen++;
3761: }
1.71 noro 3762: else
3763: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
1.146 noro 3764: if ( nd_vc ) {
1.147 noro 3765: nd_subst_vector(nd_vc,CP(t),nd_subst,&cp);
1.146 noro 3766: c = (Q)cp;
3767: } else
3768: c = CQ(t);
3769: r = rem(NM(c),mod);
1.71 noro 3770: if ( r ) {
1.146 noro 3771: if ( SGN(c) < 0 )
1.113 noro 3772: r = mod-r;
1.146 noro 3773: if ( DN(c) ) {
3774: s = rem(DN(c),mod);
1.125 noro 3775: if ( !s )
3776: error("ndv_mod : division by 0");
3777: s = invm(s,mod);
3778: DMAR(r,s,0,mod,u); r = u;
3779: }
1.71 noro 3780: CM(d) = r;
3781: ndl_copy(DL(t),DL(d));
3782: NMV_ADV(d);
3783: dlen++;
3784: }
3785: }
1.61 noro 3786: LEN(p) = dlen;
3787: }
3788:
3789: NDV ptondv(VL vl,VL dvl,P p)
3790: {
3791: ND nd;
3792:
3793: nd = ptond(vl,dvl,p);
3794: return ndtondv(0,nd);
3795: }
3796:
3797: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3798: {
1.61 noro 3799: int n,i,j,k,e;
3800: VL tvl;
3801: V v;
3802: DCP dc;
3803: DCP *w;
3804: ND r,s,t,u;
3805: P x;
3806: int c;
3807: UINT *d;
1.23 noro 3808: NM m,m0;
1.61 noro 3809:
3810: if ( !p )
3811: return 0;
3812: else if ( NUM(p) ) {
3813: NEWNM(m);
3814: ndl_zero(DL(m));
1.113 noro 3815: CQ(m) = (Q)p;
1.61 noro 3816: NEXT(m) = 0;
3817: MKND(nd_nvar,m,1,r);
3818: SG(r) = 0;
3819: return r;
3820: } else {
3821: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3822: w = (DCP *)ALLOCA(k*sizeof(DCP));
3823: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3824: for ( i = 0, tvl = dvl, v = VR(p);
1.98 noro 3825: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
1.61 noro 3826: if ( !tvl ) {
3827: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3828: t = ptond(vl,dvl,COEF(w[j]));
3829: pwrp(vl,x,DEG(w[j]),&p);
3830: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3831: }
3832: return s;
3833: } else {
3834: NEWNM(m0); d = DL(m0);
3835: for ( j = k-1, s = 0; j >= 0; j-- ) {
3836: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3837: TD(d) = MUL_WEIGHT(e,i);
3838: if ( nd_blockmask) ndl_weight_mask(d);
3839: t = ptond(vl,dvl,COEF(w[j]));
3840: for ( m = BDY(t); m; m = NEXT(m) )
3841: ndl_addto(DL(m),d);
3842: SG(t) += TD(d);
3843: s = nd_add(0,s,t);
3844: }
3845: FREENM(m0);
3846: return s;
3847: }
3848: }
3849: }
3850:
3851: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3852: {
3853: VL tvl;
3854: int len,n,j,i,e;
3855: NMV m;
3856: Q q;
3857: P c;
3858: UINT *d;
3859: P s,r,u,t,w;
1.71 noro 3860: GFS gfs;
1.23 noro 3861:
1.34 noro 3862: if ( !p ) return 0;
1.61 noro 3863: else {
3864: len = LEN(p);
3865: n = NV(p);
3866: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3867: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3868: if ( mod == -1 ) {
3869: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3870: } else if ( mod ) {
1.61 noro 3871: STOQ(CM(m),q); c = (P)q;
3872: } else
1.113 noro 3873: c = CP(m);
1.61 noro 3874: d = DL(m);
3875: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3876: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3877: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3878: }
3879: addp(vl,s,t,&u); s = u;
3880: }
3881: return s;
1.23 noro 3882: }
1.3 noro 3883: }
3884:
1.61 noro 3885: NDV ndtondv(int mod,ND p)
1.11 noro 3886: {
3887: NDV d;
1.61 noro 3888: NMV m,m0;
3889: NM t;
3890: int i,len;
1.11 noro 3891:
1.34 noro 3892: if ( !p ) return 0;
1.61 noro 3893: len = LEN(p);
1.74 noro 3894: if ( mod )
3895: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3896: else
3897: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 3898: #if 0
1.74 noro 3899: ndv_alloc += nmv_adv*len;
1.103 noro 3900: #endif
1.61 noro 3901: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3902: ndl_copy(DL(t),DL(m));
1.113 noro 3903: CQ(m) = CQ(t);
1.11 noro 3904: }
1.61 noro 3905: MKNDV(NV(p),m0,len,d);
1.14 noro 3906: SG(d) = SG(p);
1.11 noro 3907: return d;
3908: }
3909:
1.61 noro 3910: ND ndvtond(int mod,NDV p)
1.11 noro 3911: {
1.61 noro 3912: ND d;
3913: NM m,m0;
1.11 noro 3914: NMV t;
1.61 noro 3915: int i,len;
1.11 noro 3916:
1.34 noro 3917: if ( !p ) return 0;
1.11 noro 3918: m0 = 0;
1.61 noro 3919: len = p->len;
3920: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3921: NEXTNM(m0,m);
3922: ndl_copy(DL(t),DL(m));
1.113 noro 3923: CQ(m) = CQ(t);
1.11 noro 3924: }
3925: NEXT(m) = 0;
1.61 noro 3926: MKND(NV(p),m0,len,d);
1.14 noro 3927: SG(d) = SG(p);
1.11 noro 3928: return d;
3929: }
3930:
1.3 noro 3931: void ndv_print(NDV p)
3932: {
3933: NMV m;
3934: int i,len;
3935:
1.34 noro 3936: if ( !p ) printf("0\n");
1.3 noro 3937: else {
1.14 noro 3938: len = LEN(p);
1.3 noro 3939: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3940: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3941: else printf("+%d*",CM(m));
1.16 noro 3942: ndl_print(DL(m));
3943: }
3944: printf("\n");
3945: }
3946: }
3947:
1.113 noro 3948: void ndv_print_q(NDV p)
1.16 noro 3949: {
3950: NMV m;
3951: int i,len;
3952:
1.34 noro 3953: if ( !p ) printf("0\n");
1.16 noro 3954: else {
3955: len = LEN(p);
3956: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3957: printf("+");
1.113 noro 3958: printexpr(CO,(Obj)CQ(m));
1.16 noro 3959: printf("*");
1.14 noro 3960: ndl_print(DL(m));
1.3 noro 3961: }
3962: printf("\n");
3963: }
1.25 noro 3964: }
3965:
1.61 noro 3966: NODE ndv_reducebase(NODE x)
1.27 noro 3967: {
3968: int len,i,j;
3969: NDV *w;
3970: NODE t,t0;
3971:
3972: len = length(x);
3973: w = (NDV *)ALLOCA(len*sizeof(NDV));
3974: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3975: for ( i = 0; i < len; i++ ) {
3976: for ( j = 0; j < i; j++ ) {
3977: if ( w[i] && w[j] )
3978: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3979: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3980: }
3981: }
3982: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3983: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3984: }
3985: NEXT(t) = 0; x = t0;
3986: return x;
1.11 noro 3987: }
1.32 noro 3988:
1.43 noro 3989: /* XXX incomplete */
3990:
1.32 noro 3991: void nd_init_ord(struct order_spec *ord)
3992: {
1.43 noro 3993: switch ( ord->id ) {
1.32 noro 3994: case 0:
1.43 noro 3995: switch ( ord->ord.simple ) {
3996: case 0:
3997: nd_dcomp = 1;
3998: nd_isrlex = 1;
3999: break;
4000: case 1:
4001: nd_dcomp = 1;
4002: nd_isrlex = 0;
4003: break;
4004: case 2:
4005: nd_dcomp = 0;
4006: nd_isrlex = 0;
1.45 noro 4007: ndl_compare_function = ndl_lex_compare;
1.58 noro 4008: break;
4009: case 11:
4010: /* XXX */
4011: nd_dcomp = 0;
4012: nd_isrlex = 1;
4013: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 4014: break;
4015: default:
4016: error("nd_gr : unsupported order");
4017: }
1.32 noro 4018: break;
4019: case 1:
1.96 noro 4020: /* block order */
1.43 noro 4021: /* XXX */
4022: nd_dcomp = -1;
1.32 noro 4023: nd_isrlex = 0;
1.45 noro 4024: ndl_compare_function = ndl_block_compare;
1.34 noro 4025: break;
1.43 noro 4026: case 2:
1.96 noro 4027: /* matrix order */
4028: /* XXX */
4029: nd_dcomp = -1;
4030: nd_isrlex = 0;
4031: nd_matrix_len = ord->ord.matrix.row;
4032: nd_matrix = ord->ord.matrix.matrix;
4033: ndl_compare_function = ndl_matrix_compare;
4034: break;
4035: case 3:
1.97 noro 4036: /* composite order */
4037: nd_dcomp = -1;
4038: nd_isrlex = 0;
4039: nd_worb_len = ord->ord.composite.length;
4040: nd_worb = ord->ord.composite.w_or_b;
4041: ndl_compare_function = ndl_composite_compare;
1.32 noro 4042: break;
4043: }
1.41 noro 4044: nd_ord = ord;
1.32 noro 4045: }
4046:
1.43 noro 4047: BlockMask nd_create_blockmask(struct order_spec *ord)
4048: {
4049: int n,i,j,s,l;
1.61 noro 4050: UINT *t;
1.43 noro 4051: BlockMask bm;
4052:
1.96 noro 4053: /* we only create mask table for block order */
4054: if ( ord->id != 1 )
1.43 noro 4055: return 0;
4056: n = ord->ord.block.length;
4057: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
4058: bm->n = n;
4059: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 4060: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 4061: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 4062: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 4063: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
4064: l = bm->order_pair[i].length;
4065: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
4066: }
4067: return bm;
1.57 noro 4068: }
4069:
4070: EPOS nd_create_epos(struct order_spec *ord)
4071: {
4072: int i,j,l,s,ord_l,ord_o;
4073: EPOS epos;
4074: struct order_pair *op;
4075:
4076: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
4077: switch ( ord->id ) {
4078: case 0:
4079: if ( nd_isrlex ) {
4080: for ( i = 0; i < nd_nvar; i++ ) {
4081: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
4082: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
4083: }
4084: } else {
4085: for ( i = 0; i < nd_nvar; i++ ) {
4086: epos[i].i = nd_exporigin + i/nd_epw;
4087: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4088: }
4089: }
4090: break;
4091: case 1:
4092: /* block order */
4093: l = ord->ord.block.length;
4094: op = ord->ord.block.order_pair;
4095: for ( j = 0, s = 0; j < l; j++ ) {
4096: ord_o = op[j].order;
4097: ord_l = op[j].length;
4098: if ( !ord_o )
4099: for ( i = 0; i < ord_l; i++ ) {
4100: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
4101: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
4102: }
4103: else
4104: for ( i = 0; i < ord_l; i++ ) {
4105: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4106: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4107: }
4108: s += ord_l;
4109: }
4110: break;
4111: case 2:
1.96 noro 4112: /* matrix order */
4113: case 3:
4114: /* composite order */
4115: for ( i = 0; i < nd_nvar; i++ ) {
4116: epos[i].i = nd_exporigin + i/nd_epw;
4117: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4118: }
4119: break;
1.57 noro 4120: }
4121: return epos;
1.43 noro 4122: }
1.59 noro 4123:
4124: /* external interface */
4125:
4126: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4127: {
1.61 noro 4128: NODE t,in0,in;
1.59 noro 4129: ND nd,nf;
1.61 noro 4130: NDV ndv;
4131: VL vv,tv;
4132: int stat,nvar,max,e;
1.113 noro 4133: union oNDC dn;
1.146 noro 4134: Q cont;
4135: P pp;
1.59 noro 4136:
1.116 noro 4137: if ( !f ) {
4138: *rp = 0;
4139: return;
4140: }
1.59 noro 4141: pltovl(v,&vv);
1.74 noro 4142: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 4143:
4144: /* get the degree bound */
4145: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4146: for ( tv = vv; tv; tv = NEXT(tv) ) {
4147: e = getdeg(tv->v,(P)BDY(t));
4148: max = MAX(e,max);
4149: }
4150: for ( tv = vv; tv; tv = NEXT(tv) ) {
4151: e = getdeg(tv->v,f);
4152: max = MAX(e,max);
4153: }
4154:
1.59 noro 4155: nd_init_ord(ord);
1.61 noro 4156: nd_setup_parameters(nvar,max);
4157:
4158: /* conversion to ndv */
4159: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4160: NEXTNODE(in0,in);
1.146 noro 4161: ptozp((P)BDY(t),1,&cont,&pp);
4162: BDY(in) = (pointer)ptondv(CO,vv,pp);
1.69 noro 4163: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4164: }
4165: NEXTNODE(in0,in);
4166: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 4167: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 4168: NEXT(in) = 0;
4169:
1.128 noro 4170: /* dont sort */
4171: ndv_setup(m,0,in0,1);
1.59 noro 4172: nd_psn--;
4173: nd_scale=2;
4174: while ( 1 ) {
4175: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.146 noro 4176: stat = nd_nf(m,0,nd,nd_ps,1,0,&nf);
1.59 noro 4177: if ( !stat ) {
4178: nd_psn++;
1.103 noro 4179: nd_reconstruct(0,0);
1.59 noro 4180: nd_psn--;
4181: } else
4182: break;
4183: }
1.61 noro 4184: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4185: }
4186:
4187: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4188: {
4189: NM m;
4190: UINT *t,*s;
4191: int i;
4192:
4193: for ( i = 0; i < n; i++ ) r[i] = 0;
4194: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4195: t = DL(m);
4196: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4197: r[i] = CM(m);
4198: }
4199: for ( i = 0; !r[i]; i++ );
4200: return i;
4201: }
4202:
1.113 noro 4203: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4204: {
1.107 noro 4205: NM m;
1.74 noro 4206: UINT *t,*s;
1.107 noro 4207: int i;
1.74 noro 4208:
4209: for ( i = 0; i < n; i++ ) r[i] = 0;
1.107 noro 4210: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
1.74 noro 4211: t = DL(m);
4212: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.113 noro 4213: r[i] = CQ(m);
1.74 noro 4214: }
4215: for ( i = 0; !r[i]; i++ );
4216: return i;
4217: }
4218:
1.129 noro 4219: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4220: {
4221: NM m;
4222: NMV mr;
4223: UINT *d,*t,*s;
4224: NDV p;
4225: int i,j,len;
4226: Q *r;
4227:
4228: m = pair->mul;
4229: d = DL(m);
4230: p = nd_ps[pair->index];
4231: len = LEN(p);
4232: r = (Q *)CALLOC(n,sizeof(Q));
4233: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4234: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4235: ndl_add(d,DL(mr),t);
4236: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4237: r[i] = CQ(mr);
4238: }
4239: return r;
4240: }
4241:
1.67 noro 4242: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4243: {
4244: NM m;
4245: NMV mr;
4246: UINT *d,*t,*s;
4247: NDV p;
1.67 noro 4248: unsigned char *ivc;
4249: unsigned short *ivs;
1.81 noro 4250: UINT *v,*ivi,*s0v;
1.67 noro 4251: int i,j,len,prev,diff,cdiff;
4252: IndArray r;
1.64 noro 4253:
4254: m = pair->mul;
4255: d = DL(m);
4256: p = nd_ps[pair->index];
4257: len = LEN(p);
4258: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 4259: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 4260: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4261: ndl_add(d,DL(mr),t);
4262: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 4263: v[j] = i;
4264: }
1.81 noro 4265: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 4266: r->head = v[0];
4267: diff = 0;
4268: for ( i = 1; i < len; i++ ) {
4269: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4270: }
4271: if ( diff < 256 ) {
4272: r->width = 1;
4273: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4274: r->index.c = ivc;
4275: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4276: } else if ( diff < 65536 ) {
4277: r->width = 2;
4278: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4279: r->index.s = ivs;
4280: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4281: } else {
4282: r->width = 4;
4283: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4284: r->index.i = ivi;
4285: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 4286: }
1.67 noro 4287: return r;
1.64 noro 4288: }
4289:
1.135 noro 4290: int compress_array(Q *svect,Q *cvect,int n)
4291: {
4292: int i,j;
4293:
4294: for ( i = j = 0; i < n; i++ )
4295: if ( svect[i] ) cvect[j++] = svect[i];
4296: return j;
4297: }
4298:
4299: void expand_array(Q *svect,Q *cvect,int n)
4300: {
4301: int i,j;
4302:
4303: for ( i = j = 0; j < n; i++ )
4304: if ( svect[i] ) svect[i] = cvect[j++];
4305: }
4306:
1.133 noro 4307: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4308: {
1.135 noro 4309: int i,j,k,len,pos,prev,nz;
1.113 noro 4310: Q cs,mcs,c1,c2,cr,gcd,t;
1.107 noro 4311: IndArray ivect;
4312: unsigned char *ivc;
4313: unsigned short *ivs;
4314: unsigned int *ivi;
4315: NDV redv;
4316: NMV mr;
4317: NODE rp;
4318: int maxrs;
1.133 noro 4319: double hmag;
1.135 noro 4320: Q *cvect;
1.107 noro 4321:
4322: maxrs = 0;
1.134 noro 4323: for ( i = 0; i < col && !svect[i]; i++ );
4324: if ( i == col ) return maxrs;
4325: hmag = p_mag((P)svect[i])*nd_scale;
1.135 noro 4326: cvect = (Q *)ALLOCA(col*sizeof(Q));
1.107 noro 4327: for ( i = 0; i < nred; i++ ) {
4328: ivect = imat[i];
4329: k = ivect->head;
4330: if ( svect[k] ) {
4331: maxrs = MAX(maxrs,rp0[i]->sugar);
1.133 noro 4332: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
1.107 noro 4333: len = LEN(redv); mr = BDY(redv);
1.113 noro 4334: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4335: chsgnq(cs,&mcs);
4336: if ( !UNIQ(cr) ) {
4337: for ( j = 0; j < col; j++ ) {
4338: mulq(svect[j],cr,&c1); svect[j] = c1;
4339: }
4340: }
1.107 noro 4341: svect[k] = 0; prev = k;
4342: switch ( ivect->width ) {
4343: case 1:
4344: ivc = ivect->index.c;
4345: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4346: pos = prev+ivc[j]; prev = pos;
1.113 noro 4347: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4348: }
4349: break;
4350: case 2:
4351: ivs = ivect->index.s;
4352: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4353: pos = prev+ivs[j]; prev = pos;
1.113 noro 4354: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4355: }
4356: break;
4357: case 4:
4358: ivi = ivect->index.i;
4359: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4360: pos = prev+ivi[j]; prev = pos;
1.113 noro 4361: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
1.107 noro 4362: }
4363: break;
4364: }
1.134 noro 4365: for ( j = k+1; j < col && !svect[j]; j++ );
4366: if ( j == col ) break;
4367: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
1.135 noro 4368: nz = compress_array(svect,cvect,col);
1.146 noro 4369: removecont_array((P *)cvect,nz);
1.135 noro 4370: expand_array(svect,cvect,nz);
1.134 noro 4371: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
4372: }
1.107 noro 4373: }
1.133 noro 4374: }
1.135 noro 4375: nz = compress_array(svect,cvect,col);
1.146 noro 4376: removecont_array((P *)cvect,nz);
1.135 noro 4377: expand_array(svect,cvect,nz);
1.133 noro 4378: if ( DP_Print ) {
4379: fprintf(asir_out,"-"); fflush(asir_out);
1.107 noro 4380: }
4381: return maxrs;
4382: }
4383:
1.76 noro 4384: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4385: {
1.67 noro 4386: int i,j,k,len,pos,prev;
1.66 noro 4387: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 4388: IndArray ivect;
4389: unsigned char *ivc;
4390: unsigned short *ivs;
4391: unsigned int *ivi;
1.65 noro 4392: NDV redv;
1.67 noro 4393: NMV mr;
1.65 noro 4394: NODE rp;
1.76 noro 4395: int maxrs;
1.65 noro 4396:
1.76 noro 4397: maxrs = 0;
1.74 noro 4398: for ( i = 0; i < nred; i++ ) {
1.65 noro 4399: ivect = imat[i];
1.67 noro 4400: k = ivect->head; svect[k] %= m;
1.65 noro 4401: if ( c = svect[k] ) {
1.76 noro 4402: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4403: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 4404: len = LEN(redv); mr = BDY(redv);
4405: svect[k] = 0; prev = k;
4406: switch ( ivect->width ) {
4407: case 1:
4408: ivc = ivect->index.c;
4409: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4410: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4411: prev = pos;
4412: DMA(c1,c,c2,up,lo);
4413: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4414: } else svect[pos] = lo;
4415: }
4416: break;
4417: case 2:
4418: ivs = ivect->index.s;
4419: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4420: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4421: prev = pos;
4422: DMA(c1,c,c2,up,lo);
4423: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4424: } else svect[pos] = lo;
4425: }
4426: break;
4427: case 4:
4428: ivi = ivect->index.i;
4429: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4430: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4431: prev = pos;
4432: DMA(c1,c,c2,up,lo);
4433: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4434: } else svect[pos] = lo;
4435: }
4436: break;
1.65 noro 4437: }
4438: }
4439: }
1.66 noro 4440: for ( i = 0; i < col; i++ )
4441: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 4442: return maxrs;
1.65 noro 4443: }
4444:
1.76 noro 4445: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4446: {
4447: int i,j,k,len,pos,prev;
4448: UINT c,c1,c2,c3,up,lo,dmy;
4449: IndArray ivect;
4450: unsigned char *ivc;
4451: unsigned short *ivs;
4452: unsigned int *ivi;
4453: NDV redv;
4454: NMV mr;
4455: NODE rp;
1.76 noro 4456: int maxrs;
1.72 noro 4457:
1.76 noro 4458: maxrs = 0;
1.74 noro 4459: for ( i = 0; i < nred; i++ ) {
1.72 noro 4460: ivect = imat[i];
4461: k = ivect->head; svect[k] %= m;
4462: if ( c = svect[k] ) {
1.76 noro 4463: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4464: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4465: len = LEN(redv); mr = BDY(redv);
4466: svect[k] = 0; prev = k;
4467: switch ( ivect->width ) {
4468: case 1:
4469: ivc = ivect->index.c;
4470: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4471: pos = prev+ivc[j]; prev = pos;
4472: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4473: }
4474: break;
4475: case 2:
4476: ivs = ivect->index.s;
4477: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4478: pos = prev+ivs[j]; prev = pos;
4479: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4480: }
4481: break;
4482: case 4:
4483: ivi = ivect->index.i;
4484: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4485: pos = prev+ivi[j]; prev = pos;
4486: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4487: }
4488: break;
4489: }
4490: }
4491: }
1.76 noro 4492: return maxrs;
1.72 noro 4493: }
4494:
1.65 noro 4495: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4496: {
4497: int j,k,len;
4498: UINT *p;
4499: UINT c;
4500: NDV r;
4501: NMV mr0,mr;
4502:
4503: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4504: if ( !len ) return 0;
4505: else {
1.74 noro 4506: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 4507: #if 0
1.74 noro 4508: ndv_alloc += nmv_adv*len;
1.103 noro 4509: #endif
1.65 noro 4510: mr = mr0;
4511: p = s0vect;
4512: for ( j = k = 0; j < col; j++, p += nd_wpd )
4513: if ( !rhead[j] ) {
4514: if ( c = vect[k++] ) {
4515: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4516: }
4517: }
4518: MKNDV(nd_nvar,mr0,len,r);
4519: return r;
4520: }
4521: }
4522:
1.129 noro 4523: /* for preprocessed vector */
4524:
1.113 noro 4525: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 4526: {
4527: int j,k,len;
4528: UINT *p;
1.113 noro 4529: Q c;
1.107 noro 4530: NDV r;
4531: NMV mr0,mr;
4532:
4533: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4534: if ( !len ) return 0;
4535: else {
1.109 noro 4536: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 4537: #if 0
4538: ndv_alloc += nmv_adv*len;
4539: #endif
4540: mr = mr0;
4541: p = s0vect;
4542: for ( j = k = 0; j < col; j++, p += nd_wpd )
4543: if ( !rhead[j] ) {
4544: if ( c = vect[k++] ) {
1.113 noro 4545: if ( DN(c) )
4546: error("afo");
4547: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
1.107 noro 4548: }
4549: }
4550: MKNDV(nd_nvar,mr0,len,r);
4551: return r;
4552: }
4553: }
4554:
1.129 noro 4555: /* for plain vector */
4556:
4557: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
4558: {
4559: int j,k,len;
4560: UINT *p;
4561: Q c;
4562: NDV r;
4563: NMV mr0,mr;
4564:
4565: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
4566: if ( !len ) return 0;
4567: else {
4568: mr0 = (NMV)GC_malloc(nmv_adv*len);
4569: #if 0
4570: ndv_alloc += nmv_adv*len;
4571: #endif
4572: mr = mr0;
4573: p = s0vect;
4574: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
4575: if ( c = vect[k] ) {
4576: if ( DN(c) )
4577: error("afo");
4578: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
4579: }
4580: MKNDV(nd_nvar,mr0,len,r);
4581: return r;
4582: }
4583: }
4584:
1.133 noro 4585: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 4586: {
4587: ND_pairs t;
4588: NODE sp0,sp;
4589: int stat;
4590: ND spol;
4591:
4592: for ( t = l; t; t = NEXT(t) ) {
1.133 noro 4593: stat = nd_sp(m,trace,t,&spol);
1.65 noro 4594: if ( !stat ) return 0;
4595: if ( spol ) {
4596: add_pbucket_symbolic(bucket,spol);
4597: }
4598: }
1.68 noro 4599: return 1;
1.65 noro 4600: }
4601:
1.133 noro 4602: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 4603: {
4604: NODE rp0,rp;
4605: NM mul,head,s0,s;
1.76 noro 4606: int index,col,i,sugar;
1.65 noro 4607: RHist h;
4608: UINT *s0v,*p;
4609: NM_ind_pair pair;
4610: ND red;
1.133 noro 4611: NDV *ps;
1.65 noro 4612:
4613: s0 = 0; rp0 = 0; col = 0;
1.133 noro 4614: ps = trace?nd_ps_trace:nd_ps;
1.65 noro 4615: while ( 1 ) {
4616: head = remove_head_pbucket_symbolic(bucket);
4617: if ( !head ) break;
4618: if ( !s0 ) s0 = head;
4619: else NEXT(s) = head;
4620: s = head;
4621: index = ndl_find_reducer(DL(head));
4622: if ( index >= 0 ) {
4623: h = nd_psh[index];
4624: NEWNM(mul);
4625: ndl_sub(DL(head),DL(h),DL(mul));
4626: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.133 noro 4627: sugar = TD(DL(mul))+SG(ps[index]);
1.76 noro 4628: MKNM_ind_pair(pair,mul,index,sugar);
1.133 noro 4629: red = ndv_mul_nm_symbolic(mul,ps[index]);
1.65 noro 4630: add_pbucket_symbolic(bucket,nd_remove_head(red));
4631: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4632: }
4633: col++;
4634: }
1.72 noro 4635: if ( rp0 ) NEXT(rp) = 0;
4636: NEXT(s) = 0;
1.65 noro 4637: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4638: for ( i = 0, p = s0v, s = s0; i < col;
4639: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4640: *s0vect = s0v;
4641: *r = rp0;
4642: return col;
4643: }
4644:
1.69 noro 4645: NODE nd_f4(int m)
4646: {
4647: int i,nh,stat,index;
4648: NODE r,g;
4649: ND_pairs d,l,t;
4650: ND spol,red;
4651: NDV nf,redv;
4652: NM s0,s;
1.84 noro 4653: NODE rp0,srp0,nflist;
1.69 noro 4654: int nsp,nred,col,rank,len,k,j,a;
4655: UINT c;
1.74 noro 4656: UINT **spmat;
1.69 noro 4657: UINT *s0vect,*svect,*p,*v;
4658: int *colstat;
4659: IndArray *imat;
4660: int *rhead;
4661: int spcol,sprow;
4662: int sugar;
4663: PGeoBucket bucket;
4664: struct oEGT eg0,eg1,eg_f4;
4665:
1.103 noro 4666: #if 0
1.74 noro 4667: ndv_alloc = 0;
1.103 noro 4668: #endif
1.69 noro 4669: g = 0; d = 0;
4670: for ( i = 0; i < nd_psn; i++ ) {
4671: d = update_pairs(d,g,i);
4672: g = update_base(g,i);
4673: }
4674: while ( d ) {
4675: get_eg(&eg0);
4676: l = nd_minsugarp(d,&d);
4677: sugar = SG(l);
4678: bucket = create_pbucket();
1.133 noro 4679: stat = nd_sp_f4(m,0,l,bucket);
1.69 noro 4680: if ( !stat ) {
4681: for ( t = l; NEXT(t); t = NEXT(t) );
4682: NEXT(t) = d; d = l;
1.103 noro 4683: d = nd_reconstruct(0,d);
1.69 noro 4684: continue;
4685: }
1.86 noro 4686: if ( bucket->m < 0 ) continue;
1.133 noro 4687: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
1.69 noro 4688: if ( !col ) {
4689: for ( t = l; NEXT(t); t = NEXT(t) );
4690: NEXT(t) = d; d = l;
1.103 noro 4691: d = nd_reconstruct(0,d);
1.69 noro 4692: continue;
4693: }
4694: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4695: if ( DP_Print )
4696: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4697: sugar,eg_f4.exectime+eg_f4.gctime);
4698: if ( 1 )
1.133 noro 4699: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
1.74 noro 4700: else
1.133 noro 4701: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
1.69 noro 4702: /* adding new bases */
1.74 noro 4703: for ( r = nflist; r; r = NEXT(r) ) {
4704: nf = (NDV)BDY(r);
1.69 noro 4705: ndv_removecont(m,nf);
1.129 noro 4706: if ( !m && nd_nalg ) {
4707: ND nf1;
4708:
4709: nf1 = ndvtond(m,nf);
4710: nd_monic(0,&nf1);
4711: nd_removecont(m,nf1);
4712: nf = ndtondv(m,nf1);
4713: }
1.77 noro 4714: nh = ndv_newps(m,nf,0);
1.69 noro 4715: d = update_pairs(d,g,nh);
4716: g = update_base(g,nh);
4717: }
4718: }
4719: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 4720: #if 0
1.74 noro 4721: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 4722: #endif
1.69 noro 4723: return g;
4724: }
1.74 noro 4725:
1.133 noro 4726: NODE nd_f4_trace(int m)
4727: {
4728: int i,nh,stat,index;
4729: NODE r,g;
4730: ND_pairs d,l,l0,t;
4731: ND spol,red;
4732: NDV nf,redv,nfqv,nfv;
4733: NM s0,s;
4734: NODE rp0,srp0,nflist;
4735: int nsp,nred,col,rank,len,k,j,a;
4736: UINT c;
4737: UINT **spmat;
4738: UINT *s0vect,*svect,*p,*v;
4739: int *colstat;
4740: IndArray *imat;
4741: int *rhead;
4742: int spcol,sprow;
4743: int sugar;
4744: PGeoBucket bucket;
4745: struct oEGT eg0,eg1,eg_f4;
4746:
4747: g = 0; d = 0;
4748: for ( i = 0; i < nd_psn; i++ ) {
4749: d = update_pairs(d,g,i);
4750: g = update_base(g,i);
4751: }
4752: while ( d ) {
4753: get_eg(&eg0);
4754: l = nd_minsugarp(d,&d);
4755: sugar = SG(l);
4756: bucket = create_pbucket();
4757: stat = nd_sp_f4(m,0,l,bucket);
4758: if ( !stat ) {
4759: for ( t = l; NEXT(t); t = NEXT(t) );
4760: NEXT(t) = d; d = l;
4761: d = nd_reconstruct(1,d);
4762: continue;
4763: }
4764: if ( bucket->m < 0 ) continue;
4765: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
4766: if ( !col ) {
4767: for ( t = l; NEXT(t); t = NEXT(t) );
4768: NEXT(t) = d; d = l;
4769: d = nd_reconstruct(1,d);
4770: continue;
4771: }
4772: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4773: if ( DP_Print )
4774: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4775: sugar,eg_f4.exectime+eg_f4.gctime);
4776: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
4777: if ( !l0 ) continue;
4778: l = l0;
4779:
4780: /* over Q */
4781: bucket = create_pbucket();
4782: stat = nd_sp_f4(0,1,l,bucket);
4783: if ( !stat ) {
4784: for ( t = l; NEXT(t); t = NEXT(t) );
4785: NEXT(t) = d; d = l;
4786: d = nd_reconstruct(1,d);
4787: continue;
4788: }
4789: if ( bucket->m < 0 ) continue;
4790: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
4791: if ( !col ) {
4792: for ( t = l; NEXT(t); t = NEXT(t) );
4793: NEXT(t) = d; d = l;
4794: d = nd_reconstruct(1,d);
4795: continue;
4796: }
4797: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
4798: /* adding new bases */
4799: for ( r = nflist; r; r = NEXT(r) ) {
4800: nfqv = (NDV)BDY(r);
4801: ndv_removecont(0,nfqv);
4802: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
4803: if ( nd_nalg ) {
4804: ND nf1;
4805:
4806: nf1 = ndvtond(m,nfqv);
4807: nd_monic(0,&nf1);
4808: nd_removecont(0,nf1);
4809: nfqv = ndtondv(0,nf1); nd_free(nf1);
4810: }
4811: nfv = ndv_dup(0,nfqv);
4812: ndv_mod(m,nfv);
4813: ndv_removecont(m,nfv);
4814: nh = ndv_newps(0,nfv,nfqv);
4815: d = update_pairs(d,g,nh);
4816: g = update_base(g,nh);
4817: }
4818: }
4819: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
4820: #if 0
4821: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
4822: #endif
4823: return g;
4824: }
4825:
4826: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 4827: {
1.67 noro 4828: IndArray *imat;
1.106 noro 4829: int nsp,nred,i;
1.65 noro 4830: int *rhead;
1.106 noro 4831: NODE r0,rp;
1.74 noro 4832: ND_pairs sp;
4833: NM_ind_pair *rvect;
1.63 noro 4834:
1.74 noro 4835: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
1.106 noro 4836: nred = length(rp0);
1.74 noro 4837: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4838: rhead = (int *)ALLOCA(col*sizeof(int));
4839: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4840:
1.74 noro 4841: /* construction of index arrays */
4842: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4843: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4844: rvect[i] = (NM_ind_pair)BDY(rp);
4845: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4846: rhead[imat[i]->head] = 1;
4847: }
1.107 noro 4848: if ( m )
1.133 noro 4849: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
1.107 noro 4850: else
1.133 noro 4851: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
1.106 noro 4852: return r0;
4853: }
1.74 noro 4854:
1.106 noro 4855: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 4856: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 4857: {
4858: int spcol,sprow,a;
4859: int i,j,k,l,rank;
4860: NODE r0,r;
4861: ND_pairs sp;
4862: ND spol;
4863: int **spmat;
4864: UINT *svect,*v;
4865: int *colstat;
4866: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4867: int maxrs;
4868: int *spsugar;
1.133 noro 4869: ND_pairs *spactive;
1.106 noro 4870:
4871: spcol = col-nred;
4872: get_eg(&eg0);
1.74 noro 4873: /* elimination (1st step) */
4874: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4875: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4876: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.133 noro 4877: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
1.74 noro 4878: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4879: nd_sp(m,0,sp,&spol);
1.75 noro 4880: if ( !spol ) continue;
1.74 noro 4881: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4882: if ( m == -1 )
4883: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4884: else
4885: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4886: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4887: if ( i < col ) {
4888: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4889: for ( j = k = 0; j < col; j++ )
4890: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4891: spsugar[sprow] = MAX(maxrs,SG(spol));
1.133 noro 4892: if ( nz )
4893: spactive[sprow] = sp;
1.74 noro 4894: sprow++;
4895: }
1.76 noro 4896: nd_free(spol);
1.74 noro 4897: }
1.80 noro 4898: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4899: if ( DP_Print ) {
4900: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4901: fflush(asir_out);
4902: }
1.74 noro 4903: /* free index arrays */
4904: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4905:
4906: /* elimination (2nd step) */
4907: colstat = (int *)ALLOCA(spcol*sizeof(int));
4908: if ( m == -1 )
1.76 noro 4909: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4910: else
1.133 noro 4911: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
1.74 noro 4912: r0 = 0;
4913: for ( i = 0; i < rank; i++ ) {
4914: NEXTNODE(r0,r); BDY(r) =
4915: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4916: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4917: GC_free(spmat[i]);
4918: }
1.109 noro 4919: if ( r0 ) NEXT(r) = 0;
1.133 noro 4920:
1.74 noro 4921: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4922: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4923: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4924: if ( DP_Print ) {
1.80 noro 4925: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4926: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4927: nsp,nred,sprow,spcol,rank);
4928: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4929: }
1.133 noro 4930: if ( nz ) {
4931: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
4932: if ( rank > 0 ) {
4933: NEXT(spactive[rank-1]) = 0;
4934: *nz = spactive[0];
4935: } else
4936: *nz = 0;
4937: }
1.74 noro 4938: return r0;
4939: }
4940:
1.133 noro 4941: #if 1
4942: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 4943: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
4944: {
4945: int spcol,sprow,a;
4946: int i,j,k,l,rank;
4947: NODE r0,r;
4948: ND_pairs sp;
4949: ND spol;
1.113 noro 4950: Q **spmat;
4951: Q *svect,*v;
1.107 noro 4952: int *colstat;
4953: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
4954: int maxrs;
4955: int *spsugar;
1.137 noro 4956: pointer *w;
1.107 noro 4957:
4958: spcol = col-nred;
4959: get_eg(&eg0);
4960: /* elimination (1st step) */
1.113 noro 4961: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
4962: svect = (Q *)ALLOCA(col*sizeof(Q));
4963: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
1.107 noro 4964: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
1.133 noro 4965: nd_sp(0,trace,sp,&spol);
1.107 noro 4966: if ( !spol ) continue;
1.113 noro 4967: nd_to_vect_q(s0vect,col,spol,svect);
1.133 noro 4968: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
1.107 noro 4969: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4970: if ( i < col ) {
1.113 noro 4971: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
1.107 noro 4972: for ( j = k = 0; j < col; j++ )
4973: if ( !rhead[j] ) v[k++] = svect[j];
4974: spsugar[sprow] = MAX(maxrs,SG(spol));
4975: sprow++;
4976: }
1.109 noro 4977: /* nd_free(spol); */
1.107 noro 4978: }
4979: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4980: if ( DP_Print ) {
4981: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4982: fflush(asir_out);
4983: }
4984: /* free index arrays */
1.109 noro 4985: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
1.107 noro 4986:
4987: /* elimination (2nd step) */
4988: colstat = (int *)ALLOCA(spcol*sizeof(int));
1.113 noro 4989: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
1.137 noro 4990: w = (pointer *)ALLOCA(rank*sizeof(pointer));
4991: for ( i = 0; i < rank; i++ ) {
4992: w[rank-i-1] = (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
4993: SG((NDV)w[rank-i-1]) = spsugar[i];
4994: /* GC_free(spmat[i]); */
4995: }
1.138 noro 4996: #if 0
1.137 noro 4997: qsort(w,rank,sizeof(NDV),
4998: (int (*)(const void *,const void *))ndv_compare);
4999: #endif
1.107 noro 5000: r0 = 0;
5001: for ( i = 0; i < rank; i++ ) {
1.137 noro 5002: NEXTNODE(r0,r); BDY(r) = w[i];
1.107 noro 5003: }
1.137 noro 5004: if ( r0 ) NEXT(r) = 0;
1.135 noro 5005:
1.109 noro 5006: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
1.107 noro 5007: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
5008: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
5009: if ( DP_Print ) {
5010: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
5011: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
5012: nsp,nred,sprow,spcol,rank);
5013: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
5014: }
5015: return r0;
5016: }
1.129 noro 5017: #else
5018: void printm(Q **mat,int row,int col)
5019: {
5020: int i,j;
5021: printf("[");
5022: for ( i = 0; i < row; i++ ) {
5023: for ( j = 0; j < col; j++ ) {
5024: printexpr(CO,mat[i][j]); printf(" ");
5025: }
5026: printf("]\n");
5027: }
5028: }
5029:
5030: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
5031: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
5032: {
5033: int row,a;
5034: int i,j,rank;
5035: NODE r0,r;
5036: ND_pairs sp;
5037: ND spol;
5038: Q **mat;
5039: int *colstat;
5040: int *sugar;
5041:
5042: row = nsp+nred;
5043: /* make the matrix */
5044: mat = (Q **)ALLOCA(row*sizeof(Q *));
5045: sugar = (int *)ALLOCA(row*sizeof(int));
5046: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
5047: nd_sp(0,0,sp,&spol);
5048: if ( !spol ) continue;
5049: mat[row] = (Q *)MALLOC(col*sizeof(Q));
5050: nd_to_vect_q(s0vect,col,spol,mat[row]);
5051: sugar[row] = SG(spol);
5052: row++;
5053: }
5054: for ( i = 0; i < nred; i++, row++ ) {
5055: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
5056: sugar[row] = rvect[i]->sugar;
5057: }
5058: /* elimination */
5059: colstat = (int *)ALLOCA(col*sizeof(int));
5060: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
5061: r0 = 0;
5062: for ( i = 0; i < rank; i++ ) {
5063: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
5064: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
5065: if ( rhead[j] ) continue;
5066: NEXTNODE(r0,r); BDY(r) =
5067: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
5068: SG((NDV)BDY(r)) = sugar[i];
5069: }
5070: if ( r0 ) NEXT(r) = 0;
5071: printf("\n");
5072: return r0;
5073: }
5074: #endif
1.107 noro 5075:
1.74 noro 5076: FILE *nd_write,*nd_read;
5077:
5078: void nd_send_int(int a) {
5079: write_int(nd_write,&a);
5080: }
5081:
5082: void nd_send_intarray(int *p,int len) {
5083: write_intarray(nd_write,p,len);
5084: }
5085:
5086: int nd_recv_int() {
5087: int a;
5088:
5089: read_int(nd_read,&a);
5090: return a;
5091: }
5092:
5093: void nd_recv_intarray(int *p,int len) {
5094: read_intarray(nd_read,p,len);
5095: }
5096:
5097: void nd_send_ndv(NDV p) {
5098: int len,i;
5099: NMV m;
5100:
5101: if ( !p ) nd_send_int(0);
5102: else {
5103: len = LEN(p);
5104: nd_send_int(len);
5105: m = BDY(p);
5106: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5107: nd_send_int(CM(m));
5108: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 5109: }
1.74 noro 5110: }
5111: }
5112:
5113: void nd_send_nd(ND p) {
5114: int len,i;
5115: NM m;
5116:
5117: if ( !p ) nd_send_int(0);
5118: else {
5119: len = LEN(p);
5120: nd_send_int(len);
5121: m = BDY(p);
5122: for ( i = 0; i < len; i++, m = NEXT(m) ) {
5123: nd_send_int(CM(m));
5124: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 5125: }
1.74 noro 5126: }
5127: }
1.65 noro 5128:
1.74 noro 5129: NDV nd_recv_ndv()
5130: {
5131: int len,i;
5132: NMV m,m0;
5133: NDV r;
1.65 noro 5134:
1.74 noro 5135: len = nd_recv_int();
5136: if ( !len ) return 0;
5137: else {
5138: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5139: #if 0
1.74 noro 5140: ndv_alloc += len*nmv_adv;
1.103 noro 5141: #endif
1.74 noro 5142: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5143: CM(m) = nd_recv_int();
5144: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 5145: }
1.74 noro 5146: MKNDV(nd_nvar,m0,len,r);
5147: return r;
5148: }
5149: }
1.65 noro 5150:
1.74 noro 5151: int ox_exec_f4_red(Q proc)
5152: {
5153: Obj obj;
5154: STRING fname;
5155: NODE arg;
5156: int s;
5157: extern int ox_need_conv,ox_file_io;
5158:
5159: MKSTR(fname,"nd_exec_f4_red");
5160: arg = mknode(2,proc,fname);
5161: Pox_cmo_rpc(arg,&obj);
5162: s = get_ox_server_id(QTOS(proc));
5163: nd_write = iofp[s].out;
5164: nd_read = iofp[s].in;
5165: ox_need_conv = ox_file_io = 0;
5166: return s;
5167: }
5168:
1.133 noro 5169: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5170: {
5171: int nsp,nred;
5172: int i,rank,s;
5173: NODE rp,r0,r;
5174: ND_pairs sp;
5175: NM_ind_pair pair;
5176: NMV nmv;
5177: NM nm;
5178: NDV nf;
5179: Obj proc,dmy;
5180:
5181: ox_launch_main(0,0,&proc);
5182: s = ox_exec_f4_red((Q)proc);
5183:
5184: nd_send_int(m);
5185: nd_send_int(nd_nvar);
5186: nd_send_int(nd_bpe);
5187: nd_send_int(nd_wpd);
5188: nd_send_int(nmv_adv);
5189:
1.89 noro 5190: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 5191:
5192: nd_send_int(nd_psn);
5193: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5194:
5195: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5196: nd_send_int(nsp);
5197: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5198: nd_send_int(sp->i1); nd_send_int(sp->i2);
5199: }
5200:
5201: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5202:
5203: nred = length(rp0); nd_send_int(nred);
5204: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5205: pair = (NM_ind_pair)BDY(rp);
5206: nd_send_int(pair->index);
5207: nd_send_intarray(pair->mul->dl,nd_wpd);
5208: }
5209: fflush(nd_write);
5210: rank = nd_recv_int();
5211: fprintf(asir_out,"rank=%d\n",rank);
5212: r0 = 0;
5213: for ( i = 0; i < rank; i++ ) {
5214: nf = nd_recv_ndv();
5215: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5216: }
5217: Pox_shutdown(mknode(1,proc),&dmy);
5218: return r0;
5219: }
5220:
5221: /* server side */
5222:
5223: void nd_exec_f4_red_dist()
5224: {
5225: int m,i,nsp,col,s0size,nred,spcol,j,k;
5226: NM_ind_pair *rp0;
5227: NDV nf;
5228: UINT *s0vect;
5229: IndArray *imat;
5230: int *rhead;
5231: int **spmat;
5232: UINT *svect,*v;
5233: ND_pairs *sp0;
5234: int *colstat;
5235: int a,sprow,rank;
1.89 noro 5236: struct order_spec *ord;
1.74 noro 5237: Obj ordspec;
5238: ND spol;
1.76 noro 5239: int maxrs;
5240: int *spsugar;
1.74 noro 5241:
5242: nd_read = iofp[0].in;
5243: nd_write = iofp[0].out;
5244: m = nd_recv_int();
5245: nd_nvar = nd_recv_int();
5246: nd_bpe = nd_recv_int();
5247: nd_wpd = nd_recv_int();
5248: nmv_adv = nd_recv_int();
5249:
5250: loadobj(nd_read,&ordspec);
1.89 noro 5251: create_order_spec(0,ordspec,&ord);
5252: nd_init_ord(ord);
1.74 noro 5253: nd_setup_parameters(nd_nvar,0);
5254:
5255: nd_psn = nd_recv_int();
5256: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5257: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5258: for ( i = 0; i < nd_psn; i++ ) {
5259: nd_ps[i] = nd_recv_ndv();
5260: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5261: }
5262:
5263: nsp = nd_recv_int();
5264: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5265: for ( i = 0; i < nsp; i++ ) {
5266: NEWND_pairs(sp0[i]);
5267: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5268: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5269: }
5270:
5271: col = nd_recv_int();
5272: s0size = col*nd_wpd;
5273: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5274: nd_recv_intarray(s0vect,s0size);
5275:
5276: nred = nd_recv_int();
5277: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5278: for ( i = 0; i < nred; i++ ) {
5279: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5280: rp0[i]->index = nd_recv_int();
5281: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5282: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5283: }
5284:
5285: spcol = col-nred;
5286: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5287: rhead = (int *)MALLOC(col*sizeof(int));
5288: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5289:
5290: /* construction of index arrays */
5291: for ( i = 0; i < nred; i++ ) {
5292: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5293: rhead[imat[i]->head] = 1;
5294: }
5295:
5296: /* elimination (1st step) */
5297: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5298: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 5299: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 5300: for ( a = sprow = 0; a < nsp; a++ ) {
5301: nd_sp(m,0,sp0[a],&spol);
1.75 noro 5302: if ( !spol ) continue;
1.74 noro 5303: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 5304: if ( m == -1 )
5305: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5306: else
5307: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 5308: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5309: if ( i < col ) {
5310: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5311: for ( j = k = 0; j < col; j++ )
5312: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 5313: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 5314: sprow++;
5315: }
1.76 noro 5316: nd_free(spol);
1.74 noro 5317: }
5318: /* elimination (2nd step) */
5319: colstat = (int *)ALLOCA(spcol*sizeof(int));
5320: if ( m == -1 )
1.76 noro 5321: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 5322: else
1.133 noro 5323: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
1.74 noro 5324: nd_send_int(rank);
5325: for ( i = 0; i < rank; i++ ) {
5326: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5327: nd_send_ndv(nf);
1.63 noro 5328: }
1.74 noro 5329: fflush(nd_write);
1.107 noro 5330: }
5331:
1.113 noro 5332: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5333: {
1.109 noro 5334: int mod,i,j,t,c,rank,rank0,inv;
5335: int *ci,*ri;
1.113 noro 5336: Q dn;
1.109 noro 5337: MAT m,nm;
1.108 noro 5338: int **wmat;
5339:
5340: /* XXX */
5341: mod = 99999989;
5342: wmat = (int **)ALLOCA(row*sizeof(int *));
5343: for ( i = 0; i < row; i++ ) {
5344: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5345: for ( j = 0; j < col; j++ ) {
1.113 noro 5346: if ( mat0[i][j] ) {
5347: t = rem(NM(mat0[i][j]),mod);
5348: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5349: wmat[i][j] = t;
5350: } else
1.108 noro 5351: wmat[i][j] = 0;
5352: }
5353: }
1.133 noro 5354: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
1.109 noro 5355: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
1.113 noro 5356: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
1.109 noro 5357: if ( rank != rank0 )
5358: error("afo");
5359: for ( i = 0; i < row; i++ )
5360: for ( j = 0; j < col; j++ )
5361: mat0[i][j] = 0;
5362: c = col-rank;
5363: for ( i = 0; i < rank; i++ ) {
1.113 noro 5364: mat0[i][ri[i]] = dn;
1.109 noro 5365: for ( j = 0; j < c; j++ )
1.113 noro 5366: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
1.109 noro 5367: }
1.113 noro 5368: inv = invm(rem(NM(dn),mod),mod);
5369: if ( SGN(dn) < 0 ) inv = mod-inv;
1.109 noro 5370: for ( i = 0; i < row; i++ )
5371: for ( j = 0; j < col; j++ ) {
1.113 noro 5372: if ( mat0[i][j] ) {
5373: t = rem(NM(mat0[i][j]),mod);
5374: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5375: } else
1.109 noro 5376: t = 0;
5377: c = dmar(t,inv,0,mod);
5378: if ( wmat[i][j] != c )
5379: error("afo");
5380: }
5381: return rank;
1.76 noro 5382: }
5383:
1.133 noro 5384: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5385: {
5386: int i,j,k,l,inv,a,rank,s;
5387: unsigned int *t,*pivot,*pk;
5388: unsigned int **mat;
1.133 noro 5389: ND_pairs pair;
1.76 noro 5390:
5391: mat = (unsigned int **)mat0;
5392: for ( rank = 0, j = 0; j < col; j++ ) {
5393: for ( i = rank; i < row; i++ )
5394: mat[i][j] %= md;
5395: for ( i = rank; i < row; i++ )
5396: if ( mat[i][j] )
5397: break;
5398: if ( i == row ) {
5399: colstat[j] = 0;
5400: continue;
5401: } else
5402: colstat[j] = 1;
5403: if ( i != rank ) {
5404: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5405: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
1.133 noro 5406: if ( spactive ) {
5407: pair = spactive[i]; spactive[i] = spactive[rank];
5408: spactive[rank] = pair;
5409: }
1.76 noro 5410: }
5411: pivot = mat[rank];
5412: s = sugar[rank];
5413: inv = invm(pivot[j],md);
5414: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5415: if ( *pk ) {
5416: if ( *pk >= (unsigned int)md )
5417: *pk %= md;
5418: DMAR(*pk,inv,0,md,*pk)
5419: }
5420: for ( i = rank+1; i < row; i++ ) {
5421: t = mat[i];
5422: if ( a = t[j] ) {
5423: sugar[i] = MAX(sugar[i],s);
5424: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5425: }
5426: }
5427: rank++;
5428: }
5429: for ( j = col-1, l = rank-1; j >= 0; j-- )
5430: if ( colstat[j] ) {
5431: pivot = mat[l];
5432: s = sugar[l];
5433: for ( i = 0; i < l; i++ ) {
5434: t = mat[i];
5435: t[j] %= md;
5436: if ( a = t[j] ) {
5437: sugar[i] = MAX(sugar[i],s);
5438: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5439: }
5440: }
5441: l--;
5442: }
5443: for ( j = 0, l = 0; l < rank; j++ )
5444: if ( colstat[j] ) {
5445: t = mat[l];
5446: for ( k = j; k < col; k++ )
5447: if ( t[k] >= (unsigned int)md )
5448: t[k] %= md;
5449: l++;
5450: }
5451: return rank;
5452: }
5453:
5454: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5455: {
5456: int i,j,k,l,inv,a,rank,s;
5457: unsigned int *t,*pivot,*pk;
5458: unsigned int **mat;
5459:
5460: mat = (unsigned int **)mat0;
5461: for ( rank = 0, j = 0; j < col; j++ ) {
5462: for ( i = rank; i < row; i++ )
5463: if ( mat[i][j] )
5464: break;
5465: if ( i == row ) {
5466: colstat[j] = 0;
5467: continue;
5468: } else
5469: colstat[j] = 1;
5470: if ( i != rank ) {
5471: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5472: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5473: }
5474: pivot = mat[rank];
5475: s = sugar[rank];
5476: inv = _invsf(pivot[j]);
5477: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5478: if ( *pk )
5479: *pk = _mulsf(*pk,inv);
5480: for ( i = rank+1; i < row; i++ ) {
5481: t = mat[i];
5482: if ( a = t[j] ) {
5483: sugar[i] = MAX(sugar[i],s);
5484: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5485: }
5486: }
5487: rank++;
5488: }
5489: for ( j = col-1, l = rank-1; j >= 0; j-- )
5490: if ( colstat[j] ) {
5491: pivot = mat[l];
5492: s = sugar[l];
5493: for ( i = 0; i < l; i++ ) {
5494: t = mat[i];
5495: if ( a = t[j] ) {
5496: sugar[i] = MAX(sugar[i],s);
5497: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
5498: }
5499: }
5500: l--;
5501: }
5502: return rank;
1.77 noro 5503: }
5504:
5505: int ndv_ishomo(NDV p)
5506: {
5507: NMV m;
5508: int len,h;
5509:
5510: if ( !p ) return 1;
5511: len = LEN(p);
5512: m = BDY(p);
5513: h = TD(DL(m));
5514: NMV_ADV(m);
5515: for ( len--; len; len--, NMV_ADV(m) )
5516: if ( TD(DL(m)) != h ) return 0;
5517: return 1;
5518: }
5519:
5520: void ndv_save(NDV p,int index)
5521: {
5522: FILE *s;
5523: char name[BUFSIZ];
5524: short id;
5525: int nv,sugar,len,n,i,td,e,j;
5526: NMV m;
5527: unsigned int *dl;
5528:
5529: sprintf(name,"%s/%d",Demand,index);
5530: s = fopen(name,"w");
5531: savevl(s,0);
5532: if ( !p ) {
5533: saveobj(s,0);
5534: return;
5535: }
5536: id = O_DP;
5537: nv = NV(p);
5538: sugar = SG(p);
5539: len = LEN(p);
5540: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
5541: write_int(s,&len);
5542:
5543: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5544: saveobj(s,(Obj)CQ(m));
1.77 noro 5545: dl = DL(m);
5546: td = TD(dl);
5547: write_int(s,&td);
5548: for ( j = 0; j < nv; j++ ) {
5549: e = GET_EXP(dl,j);
5550: write_int(s,&e);
5551: }
5552: }
5553: fclose(s);
5554: }
5555:
5556: NDV ndv_load(int index)
5557: {
5558: FILE *s;
5559: char name[BUFSIZ];
5560: short id;
5561: int nv,sugar,len,n,i,td,e,j;
5562: NDV d;
5563: NMV m0,m;
5564: unsigned int *dl;
5565: Obj obj;
5566:
5567: sprintf(name,"%s/%d",Demand,index);
5568: s = fopen(name,"r");
5569: if ( !s ) return 0;
5570:
5571: skipvl(s);
5572: read_short(s,&id);
5573: if ( !id ) return 0;
5574: read_int(s,&nv);
5575: read_int(s,&sugar);
5576: read_int(s,&len);
5577:
5578: m0 = m = MALLOC(len*nmv_adv);
5579: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
1.113 noro 5580: loadobj(s,&obj); CQ(m) = (Q)obj;
1.77 noro 5581: dl = DL(m);
5582: ndl_zero(dl);
5583: read_int(s,&td); TD(dl) = td;
5584: for ( j = 0; j < nv; j++ ) {
5585: read_int(s,&e);
5586: PUT_EXP(dl,j,e);
5587: }
5588: if ( nd_blockmask ) ndl_weight_mask(dl);
5589: }
5590: fclose(s);
5591: MKNDV(nv,m0,len,d);
5592: SG(d) = sugar;
5593: return d;
1.99 noro 5594: }
5595:
1.102 noro 5596: void nd_det(int mod,MAT f,P *rp)
1.99 noro 5597: {
5598: VL fv,tv;
1.102 noro 5599: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
1.99 noro 5600: pointer **m;
1.113 noro 5601: Q mone;
1.125 noro 5602: P **w;
1.126 noro 5603: P mp,r;
1.102 noro 5604: NDV **dm;
5605: NDV *t,*mi,*mj;
5606: NDV d,s,mij,mjj;
5607: ND u;
5608: NMV nmv;
1.114 noro 5609: UINT *bound;
1.99 noro 5610: PGeoBucket bucket;
5611: struct order_spec *ord;
1.126 noro 5612: Q dq,dt,ds;
5613: N gn,qn,dn0,nm,dn;
1.99 noro 5614:
5615: create_order_spec(0,0,&ord);
5616: nd_init_ord(ord);
5617: get_vars((Obj)f,&fv);
5618: if ( f->row != f->col )
5619: error("nd_det : non-square matrix");
5620: n = f->row;
1.125 noro 5621: m = f->body;
1.99 noro 5622: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
1.125 noro 5623:
5624: if ( !nvar ) {
5625: if ( !mod )
5626: detp(CO,(P **)m,n,rp);
5627: else {
5628: w = (P **)almat_pointer(n,n);
5629: for ( i = 0; i < n; i++ )
5630: for ( j = 0; j < n; j++ )
5631: ptomp(mod,(P)m[i][j],&w[i][j]);
5632: detmp(CO,mod,w,n,&mp);
5633: mptop(mp,rp);
5634: }
5635: return;
5636: }
1.126 noro 5637:
5638: if ( !mod ) {
5639: w = (P **)almat_pointer(n,n);
5640: dq = ONE;
5641: for ( i = 0; i < n; i++ ) {
5642: dn0 = ONEN;
5643: for ( j = 0; j < n; j++ ) {
5644: if ( !m[i][j] ) continue;
5645: lgp(m[i][j],&nm,&dn);
5646: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
5647: }
5648: if ( !UNIN(dn0) ) {
5649: NTOQ(dn0,1,ds);
5650: for ( j = 0; j < n; j++ )
5651: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
5652: mulq(dq,ds,&dt); dq = dt;
5653: } else
5654: for ( j = 0; j < n; j++ )
5655: w[i][j] = (P)m[i][j];
5656: }
5657: m = (pointer **)w;
5658: }
1.125 noro 5659:
1.99 noro 5660: for ( i = 0, max = 0; i < n; i++ )
5661: for ( j = 0; j < n; j++ )
5662: for ( tv = fv; tv; tv = NEXT(tv) ) {
5663: e = getdeg(tv->v,(P)m[i][j]);
5664: max = MAX(e,max);
5665: }
1.114 noro 5666: nd_setup_parameters(nvar,max);
1.102 noro 5667: dm = (NDV **)almat_pointer(n,n);
1.99 noro 5668: for ( i = 0, max = 0; i < n; i++ )
1.102 noro 5669: for ( j = 0; j < n; j++ ) {
5670: dm[i][j] = ptondv(CO,fv,m[i][j]);
5671: if ( mod ) ndv_mod(mod,dm[i][j]);
5672: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
5673: }
5674: d = ptondv(CO,fv,(P)ONE);
5675: if ( mod ) ndv_mod(mod,d);
1.113 noro 5676: chsgnq(ONE,&mone);
1.99 noro 5677: for ( j = 0, sgn = 1; j < n; j++ ) {
1.125 noro 5678: if ( DP_Print ) fprintf(stderr,".",j);
1.99 noro 5679: for ( i = j; i < n && !dm[i][j]; i++ );
5680: if ( i == n ) {
5681: *rp = 0;
5682: return;
5683: }
1.102 noro 5684: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
1.99 noro 5685: for ( k = j; k < n; k++ )
5686: for ( l = j; l < n; l++ )
1.102 noro 5687: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
5688: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
1.99 noro 5689: }
5690: if ( k0 != j ) {
5691: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
5692: sgn = -sgn;
5693: }
5694: if ( l0 != j ) {
5695: for ( k = j; k < n; k++ ) {
5696: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
5697: }
5698: sgn = -sgn;
5699: }
1.114 noro 5700: bound = nd_det_compute_bound(dm,n,j);
5701: if ( ndl_check_bound(bound,bound) )
5702: nd_det_reconstruct(dm,n,j,d);
5703:
1.99 noro 5704: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
1.114 noro 5705: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
1.99 noro 5706: mi = dm[i]; mij = mi[j];
1.102 noro 5707: if ( mod )
5708: ndv_mul_c(mod,mij,mod-1);
5709: else
1.113 noro 5710: ndv_mul_c_q(mij,mone);
1.99 noro 5711: for ( k = j+1; k < n; k++ ) {
1.114 noro 5712: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
1.99 noro 5713: bucket = create_pbucket();
1.104 noro 5714: if ( mi[k] ) {
1.102 noro 5715: nmv = BDY(mjj); len = LEN(mjj);
5716: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5717: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
5718: add_pbucket(mod,bucket,u);
1.99 noro 5719: }
1.104 noro 5720: }
1.99 noro 5721: if ( mj[k] && mij ) {
1.102 noro 5722: nmv = BDY(mij); len = LEN(mij);
5723: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
5724: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
5725: add_pbucket(mod,bucket,u);
1.99 noro 5726: }
5727: }
1.104 noro 5728: u = nd_quo(mod,bucket,d);
1.102 noro 5729: mi[k] = ndtondv(mod,u);
1.99 noro 5730: }
1.114 noro 5731: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
1.99 noro 5732: }
5733: d = mjj;
5734: }
1.125 noro 5735: if ( DP_Print ) fprintf(stderr,"\n",k);
1.100 noro 5736: if ( sgn < 0 )
1.102 noro 5737: if ( mod )
5738: ndv_mul_c(mod,d,mod-1);
5739: else
1.113 noro 5740: ndv_mul_c_q(d,mone);
1.126 noro 5741: r = ndvtop(mod,CO,fv,d);
5742: if ( !mod && !UNIQ(dq) )
5743: divsp(CO,r,(P)dq,rp);
5744: else
5745: *rp = r;
1.99 noro 5746: }
5747:
1.102 noro 5748: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 5749: {
5750: NM mr,mr0;
1.102 noro 5751: NM tnm;
5752: NMV m;
1.99 noro 5753: UINT *d0,*dt,*dm;
5754: int c,n,td,i,c1,c2,len;
1.113 noro 5755: Q q;
1.99 noro 5756: ND r;
5757:
5758: if ( !p ) return 0;
5759: else {
1.102 noro 5760: n = NV(p); m = BDY(p); len = LEN(p);
1.99 noro 5761: d0 = DL(m0);
1.102 noro 5762: td = TD(d);
1.99 noro 5763: mr0 = 0;
5764: NEWNM(tnm);
1.102 noro 5765: if ( mod ) {
5766: c = CM(m0);
5767: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5768: ndl_add(DL(m),d0,DL(tnm));
5769: if ( ndl_reducible(DL(tnm),d) ) {
5770: NEXTNM(mr0,mr);
5771: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
1.104 noro 5772: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5773: }
5774: }
5775: } else {
1.113 noro 5776: q = CQ(m0);
1.102 noro 5777: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5778: ndl_add(DL(m),d0,DL(tnm));
5779: if ( ndl_reducible(DL(tnm),d) ) {
5780: NEXTNM(mr0,mr);
1.113 noro 5781: mulq(CQ(m),q,&CQ(mr));
1.104 noro 5782: ndl_copy(DL(tnm),DL(mr));
1.102 noro 5783: }
1.99 noro 5784: }
5785: }
5786: if ( !mr0 )
5787: return 0;
5788: else {
5789: NEXT(mr) = 0;
1.105 noro 5790: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
1.99 noro 5791: MKND(NV(p),mr0,len,r);
5792: SG(r) = SG(p) + TD(d0);
5793: return r;
5794: }
5795: }
1.114 noro 5796: }
5797:
5798: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
5799: {
5800: int i,obpe,oadv,h,k,l;
5801: static NM prev_nm_free_list;
5802: EPOS oepos;
5803:
5804: obpe = nd_bpe;
5805: oadv = nmv_adv;
5806: oepos = nd_epos;
5807: if ( obpe < 2 ) nd_bpe = 2;
5808: else if ( obpe < 3 ) nd_bpe = 3;
5809: else if ( obpe < 4 ) nd_bpe = 4;
5810: else if ( obpe < 5 ) nd_bpe = 5;
5811: else if ( obpe < 6 ) nd_bpe = 6;
5812: else if ( obpe < 8 ) nd_bpe = 8;
5813: else if ( obpe < 10 ) nd_bpe = 10;
5814: else if ( obpe < 16 ) nd_bpe = 16;
5815: else if ( obpe < 32 ) nd_bpe = 32;
5816: else error("nd_det_reconstruct : exponent too large");
5817:
5818: nd_setup_parameters(nd_nvar,0);
5819: prev_nm_free_list = _nm_free_list;
5820: _nm_free_list = 0;
5821: for ( k = j; k < n; k++ )
5822: for (l = j; l < n; l++ )
5823: ndv_realloc(dm[k][l],obpe,oadv,oepos);
5824: ndv_realloc(d,obpe,oadv,oepos);
5825: prev_nm_free_list = 0;
5826: #if 0
5827: GC_gcollect();
5828: #endif
5829: }
5830:
5831: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
5832: {
5833: UINT *d0,*d1,*d,*t,*r;
5834: int k,l;
5835:
5836: d0 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5837: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5838: for ( k = 0; k < nd_wpd; k++ ) d0[k] = 0;
5839: for ( k = j; k < n; k++ )
5840: for ( l = j; l < n; l++ )
5841: if ( dm[k][l] ) {
5842: d = ndv_compute_bound(dm[k][l]);
5843: ndl_lcm(d,d0,d1);
5844: t = d1; d1 = d0; d0 = t;
5845: }
5846: r = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
5847: for ( k = 0; k < nd_wpd; k++ ) r[k] = d0[k];
5848: return r;
1.117 noro 5849: }
5850:
5851: DL nd_separate_d(UINT *d,UINT *trans)
5852: {
1.121 noro 5853: int n,td,i,e,j;
1.117 noro 5854: DL a;
5855:
5856: ndl_zero(trans);
5857: td = 0;
1.121 noro 5858: for ( i = 0; i < nd_ntrans; i++ ) {
5859: e = GET_EXP(d,i);
5860: PUT_EXP(trans,i,e);
5861: td += MUL_WEIGHT(e,i);
5862: }
5863: if ( nd_ntrans+nd_nalg < nd_nvar ) {
5864: /* homogenized */
5865: i = nd_nvar-1;
1.117 noro 5866: e = GET_EXP(d,i);
5867: PUT_EXP(trans,i,e);
5868: td += MUL_WEIGHT(e,i);
5869: }
5870: TD(trans) = td;
5871: if ( nd_blockmask) ndl_weight_mask(trans);
5872: NEWDL(a,nd_nalg);
5873: td = 0;
1.121 noro 5874: for ( i = 0; i < nd_nalg; i++ ) {
5875: j = nd_ntrans+i;
5876: e = GET_EXP(d,j);
5877: a->d[i] = e;
1.117 noro 5878: td += e;
5879: }
5880: a->td = td;
5881: return a;
5882: }
5883:
1.118 noro 5884: int nd_monic(int mod,ND *p)
1.117 noro 5885: {
5886: UINT *trans,*t;
5887: DL alg;
5888: MP mp0,mp;
1.118 noro 5889: NM m,m0,m1,ma0,ma,mb,mr0,mr;
5890: ND r;
1.117 noro 5891: DL dl;
5892: DP nm;
5893: NDV ndv;
1.118 noro 5894: DAlg inv,cd;
1.117 noro 5895: ND s,c;
1.118 noro 5896: Q l,mul;
5897: N ln;
5898: int n,ntrans,i,e,td,is_lc,len;
5899: NumberField nf;
5900: struct oEGT eg0,eg1;
5901:
5902: if ( !(nf = get_numberfield()) )
5903: error("nd_monic : current_numberfield is not set");
1.117 noro 5904:
1.118 noro 5905: /* Q coef -> DAlg coef */
5906: NEWNM(ma0); ma = ma0;
5907: m = BDY(*p);
5908: is_lc = 1;
5909: while ( 1 ) {
5910: NEWMP(mp0); mp = mp0;
5911: mp->c = (P)CQ(m);
5912: mp->dl = nd_separate_d(DL(m),DL(ma));
5913: NEWNM(mb);
5914: for ( m = NEXT(m); m; m = NEXT(m) ) {
5915: alg = nd_separate_d(DL(m),DL(mb));
5916: if ( !ndl_equal(DL(ma),DL(mb)) )
5917: break;
5918: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
5919: }
5920: NEXT(mp) = 0;
5921: MKDP(nd_nalg,mp0,nm);
5922: MKDAlg(nm,ONE,cd);
5923: if ( is_lc == 1 ) {
5924: /* if the lc is a rational number, we have nothing to do */
5925: if ( !mp0->dl->td )
5926: return 1;
5927:
5928: get_eg(&eg0);
5929: invdalg(cd,&inv);
5930: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
5931: /* check the validity of inv */
5932: if ( mod && !rem(NM(inv->dn),mod) )
5933: return 0;
5934: CA(ma) = nf->one;
5935: is_lc = 0;
5936: ln = ONEN;
5937: } else {
5938: muldalg(cd,inv,&CA(ma));
5939: lcmn(ln,NM(CA(ma)->dn),&ln);
5940: }
5941: if ( m ) {
5942: NEXT(ma) = mb; ma = mb;
5943: } else {
5944: NEXT(ma) = 0;
5945: break;
5946: }
5947: }
5948: /* l = lcm(denoms) */
5949: NTOQ(ln,1,l);
5950: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
5951: divq(l,CA(m)->dn,&mul);
5952: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
5953: NEXTNM(mr0,mr);
5954: mulq((Q)mp->c,mul,&CQ(mr));
5955: dl = mp->dl;
5956: td = TD(DL(m));
5957: ndl_copy(DL(m),DL(mr));
1.121 noro 5958: for ( i = 0; i < nd_nalg; i++ ) {
5959: e = dl->d[i];
5960: PUT_EXP(DL(mr),i+nd_ntrans,e);
5961: td += MUL_WEIGHT(e,i+nd_ntrans);
1.118 noro 5962: }
5963: TD(DL(mr)) = td;
5964: if ( nd_blockmask) ndl_weight_mask(DL(mr));
5965: }
1.117 noro 5966: }
1.118 noro 5967: NEXT(mr) = 0;
5968: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
5969: MKND(NV(*p),mr0,len,r);
5970: /* XXX */
5971: SG(r) = SG(*p);
5972: nd_free(*p);
5973: *p = r;
5974: return 1;
1.59 noro 5975: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>