Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.96
1.96 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.95 2004/03/15 08:44:52 noro Exp $ */
1.2 noro 2:
1.94 noro 3: #include "nd.h"
1.63 noro 4:
1.61 noro 5: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.94 noro 6: int nd_dcomp;
7: NM _nm_free_list;
8: ND _nd_free_list;
9: ND_pairs _ndp_free_list;
1.32 noro 10:
1.74 noro 11: static int ndv_alloc;
1.87 noro 12: #if 1
1.69 noro 13: static int nd_f4_nsp=0x7fffffff;
1.87 noro 14: #else
15: static int nd_f4_nsp=50;
16: #endif
1.42 noro 17: static double nd_scale=2;
1.61 noro 18: static UINT **nd_bound;
1.42 noro 19: static struct order_spec *nd_ord;
20: static EPOS nd_epos;
1.43 noro 21: static BlockMask nd_blockmask;
1.42 noro 22: static int nd_nvar;
23: static int nd_isrlex;
24: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 25: static UINT nd_mask[32];
26: static UINT nd_mask0,nd_mask1;
1.42 noro 27:
1.20 noro 28: static NDV *nd_ps;
1.53 noro 29: static NDV *nd_ps_trace;
1.42 noro 30: static RHist *nd_psh;
31: static int nd_psn,nd_pslen;
32: static RHist *nd_red;
1.96 ! noro 33: static int *nd_work_vector;
! 34: static int **nd_matrix;
! 35: static int nd_matrix_len;
1.1 noro 36:
1.42 noro 37: static int nd_found,nd_create,nd_notfirst;
38: static int nmv_adv;
1.77 noro 39: static int nd_demand;
1.1 noro 40:
41: void nd_free_private_storage()
42: {
43: _nm_free_list = 0;
1.5 noro 44: _ndp_free_list = 0;
1.71 noro 45: #if 0
1.1 noro 46: GC_gcollect();
1.71 noro 47: #endif
1.1 noro 48: }
49:
50: void _NM_alloc()
51: {
52: NM p;
53: int i;
54:
1.11 noro 55: for ( i = 0; i < 1024; i++ ) {
1.61 noro 56: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 57: p->next = _nm_free_list; _nm_free_list = p;
58: }
59: }
60:
61: void _ND_alloc()
62: {
63: ND p;
64: int i;
65:
66: for ( i = 0; i < 1024; i++ ) {
67: p = (ND)GC_malloc(sizeof(struct oND));
68: p->body = (NM)_nd_free_list; _nd_free_list = p;
69: }
70: }
71:
72: void _NDP_alloc()
73: {
74: ND_pairs p;
75: int i;
76:
1.11 noro 77: for ( i = 0; i < 1024; i++ ) {
1.1 noro 78: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 79: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 80: p->next = _ndp_free_list; _ndp_free_list = p;
81: }
82: }
83:
1.30 noro 84: INLINE int nd_length(ND p)
1.1 noro 85: {
86: NM m;
87: int i;
88:
89: if ( !p )
90: return 0;
91: else {
92: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
93: return i;
94: }
95: }
96:
1.61 noro 97: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 98: {
1.61 noro 99: UINT u1,u2;
1.1 noro 100: int i,j;
101:
1.34 noro 102: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 103: #if USE_UNROLL
1.1 noro 104: switch ( nd_bpe ) {
1.62 noro 105: case 3:
106: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
107: u1 = d1[i]; u2 = d2[i];
108: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
109: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
110: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
111: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
112: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
113: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
114: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
115: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
116: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
117: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
118: }
119: return 1;
120: break;
1.1 noro 121: case 4:
1.41 noro 122: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 123: u1 = d1[i]; u2 = d2[i];
124: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 125: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
126: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
127: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
128: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
129: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
130: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
131: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 132: }
133: return 1;
134: break;
135: case 6:
1.41 noro 136: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 137: u1 = d1[i]; u2 = d2[i];
138: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 139: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
140: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
141: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
142: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 143: }
144: return 1;
145: break;
146: case 8:
1.41 noro 147: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 148: u1 = d1[i]; u2 = d2[i];
149: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 150: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
151: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
152: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 153: }
154: return 1;
155: break;
156: case 16:
1.41 noro 157: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 158: u1 = d1[i]; u2 = d2[i];
159: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 160: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 161: }
162: return 1;
163: break;
164: case 32:
1.41 noro 165: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 166: if ( d1[i] < d2[i] ) return 0;
167: return 1;
168: break;
169: default:
1.41 noro 170: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 171: u1 = d1[i]; u2 = d2[i];
172: for ( j = 0; j < nd_epw; j++ )
173: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
174: }
175: return 1;
176: }
1.65 noro 177: #else
178: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
179: u1 = d1[i]; u2 = d2[i];
180: for ( j = 0; j < nd_epw; j++ )
181: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
182: }
183: return 1;
184: #endif
1.1 noro 185: }
186:
1.61 noro 187: /*
188: * If the current order is a block order,
189: * then the last block is length 1 and contains
190: * the homo variable. Otherwise, the original
191: * order is either 0 or 2.
192: */
193:
194: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 195: {
1.61 noro 196: int w,i,e,n,omask0;
197:
198: omask0 = (1<<obpe)-1;
199: n = nd_nvar-1;
200: ndl_zero(r);
201: for ( i = 0; i < n; i++ ) {
202: e = GET_EXP_OLD(d,i);
203: PUT_EXP(r,i,e);
204: }
205: w = TD(d);
206: PUT_EXP(r,nd_nvar-1,weight-w);
207: TD(r) = weight;
208: if ( nd_blockmask ) ndl_weight_mask(r);
209: }
210:
211: void ndl_dehomogenize(UINT *d)
212: {
213: UINT mask;
214: UINT h;
1.31 noro 215: int i,bits;
1.23 noro 216:
1.44 noro 217: if ( nd_blockmask ) {
218: h = GET_EXP(d,nd_nvar-1);
1.45 noro 219: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 220: TD(d) -= h;
221: d[nd_exporigin-1] -= h;
222: } else {
223: if ( nd_isrlex ) {
224: if ( nd_bpe == 32 ) {
225: h = d[nd_exporigin];
226: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
227: d[i-1] = d[i];
228: d[i-1] = 0;
229: TD(d) -= h;
230: } else {
231: bits = nd_epw*nd_bpe;
232: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
233: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
234: for ( i = nd_exporigin; i < nd_wpd; i++ )
235: d[i] = ((d[i]<<nd_bpe)&mask)
236: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
237: TD(d) -= h;
238: }
1.45 noro 239: } else {
240: h = GET_EXP(d,nd_nvar-1);
241: XOR_EXP(d,nd_nvar-1,h);
242: TD(d) -= h;
243: }
1.44 noro 244: }
1.23 noro 245: }
246:
1.61 noro 247: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 248: {
1.61 noro 249: UINT t1,t2,u,u1,u2;
1.43 noro 250: int i,j,l;
1.1 noro 251:
1.65 noro 252: #if USE_UNROLL
1.1 noro 253: switch ( nd_bpe ) {
1.62 noro 254: case 3:
255: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
256: u1 = d1[i]; u2 = d2[i];
257: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
258: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
259: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
260: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
261: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
262: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
263: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
264: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
265: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
266: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
267: d[i] = u;
268: }
269: break;
1.1 noro 270: case 4:
1.41 noro 271: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 272: u1 = d1[i]; u2 = d2[i];
273: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 274: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
275: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
276: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
277: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
278: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
279: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
280: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 281: d[i] = u;
282: }
283: break;
284: case 6:
1.41 noro 285: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 286: u1 = d1[i]; u2 = d2[i];
287: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 288: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
289: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
290: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
291: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 292: d[i] = u;
293: }
294: break;
295: case 8:
1.41 noro 296: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 297: u1 = d1[i]; u2 = d2[i];
298: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 299: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
300: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
301: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 302: d[i] = u;
303: }
304: break;
305: case 16:
1.41 noro 306: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 307: u1 = d1[i]; u2 = d2[i];
308: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 309: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 310: d[i] = u;
311: }
312: break;
313: case 32:
1.41 noro 314: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 315: u1 = d1[i]; u2 = d2[i];
316: d[i] = u1>u2?u1:u2;
317: }
318: break;
319: default:
1.41 noro 320: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 321: u1 = d1[i]; u2 = d2[i];
322: for ( j = 0, u = 0; j < nd_epw; j++ ) {
323: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
324: }
325: d[i] = u;
326: }
327: break;
328: }
1.65 noro 329: #else
330: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
331: u1 = d1[i]; u2 = d2[i];
332: for ( j = 0, u = 0; j < nd_epw; j++ ) {
333: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
334: }
335: d[i] = u;
336: }
337: #endif
1.39 noro 338: TD(d) = ndl_weight(d);
1.61 noro 339: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 340: }
341:
1.61 noro 342: int ndl_weight(UINT *d)
1.1 noro 343: {
1.61 noro 344: UINT t,u;
1.1 noro 345: int i,j;
346:
1.60 noro 347: if ( current_dl_weight_vector )
348: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
349: u = GET_EXP(d,i);
350: t += MUL_WEIGHT(u,i);
351: }
352: else
353: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
354: u = d[i];
355: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
356: t += (u&nd_mask0);
357: }
1.1 noro 358: return t;
359: }
360:
1.61 noro 361: void ndl_weight_mask(UINT *d)
1.43 noro 362: {
1.61 noro 363: UINT t,u;
364: UINT *mask;
365: int i,j,k,l;
1.43 noro 366:
1.61 noro 367: l = nd_blockmask->n;
368: for ( k = 0; k < l; k++ ) {
369: mask = nd_blockmask->mask[k];
370: if ( current_dl_weight_vector )
371: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
372: u = GET_EXP_MASK(d,i,mask);
373: t += MUL_WEIGHT(u,i);
374: }
375: else
376: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
377: u = d[i]&mask[i];
378: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
379: t += (u&nd_mask0);
380: }
381: d[k+1] = t;
382: }
1.43 noro 383: }
384:
1.61 noro 385: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 386: {
387: int i;
388:
1.41 noro 389: d1 += nd_exporigin;
390: d2 += nd_exporigin;
391: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 392: if ( *d1 > *d2 )
1.32 noro 393: return nd_isrlex ? -1 : 1;
1.1 noro 394: else if ( *d1 < *d2 )
1.32 noro 395: return nd_isrlex ? 1 : -1;
1.1 noro 396: return 0;
397: }
398:
1.61 noro 399: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 400: {
401: int i,l,j,ord_o,ord_l;
402: struct order_pair *op;
1.61 noro 403: UINT t1,t2,m;
404: UINT *mask;
1.43 noro 405:
406: l = nd_blockmask->n;
407: op = nd_blockmask->order_pair;
408: for ( j = 0; j < l; j++ ) {
409: mask = nd_blockmask->mask[j];
410: ord_o = op[j].order;
411: if ( ord_o < 2 )
1.44 noro 412: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
413: else if ( t1 < t2 ) return -1;
1.43 noro 414: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 415: m = mask[i];
416: t1 = d1[i]&m;
417: t2 = d2[i]&m;
1.43 noro 418: if ( t1 > t2 )
419: return !ord_o ? -1 : 1;
420: else if ( t1 < t2 )
421: return !ord_o ? 1 : -1;
422: }
423: }
424: return 0;
425: }
426:
1.96 ! noro 427: int ndl_matrix_compare(UINT *d1,UINT *d2)
! 428: {
! 429: int i,j,s;
! 430: int *v;
! 431:
! 432: for ( j = 0; j < nd_nvar; j++ )
! 433: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
! 434: for ( i = 0; i < nd_matrix_len; i++ ) {
! 435: v = nd_matrix[i];
! 436: for ( j = 0, s = 0; j < nd_nvar; j++ )
! 437: s += v[j]*nd_work_vector[j];
! 438: if ( s > 0 ) return 1;
! 439: else if ( s < 0 ) return -1;
! 440: }
! 441: return 0;
! 442: }
! 443:
1.58 noro 444: /* TDH -> WW -> TD-> RL */
445:
1.61 noro 446: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 447: {
448: int i,m,e1,e2;
449:
450: if ( TD(d1) > TD(d2) ) return 1;
451: else if ( TD(d1) < TD(d2) ) return -1;
452: m = nd_nvar>>1;
453: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
454: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
455: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
456: }
457: if ( e1 > e2 ) return 1;
458: else if ( e1 < e2 ) return -1;
459: return ndl_lex_compare(d1,d2);
460: }
461:
1.61 noro 462: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 463: {
464: int i;
465:
1.81 noro 466: switch ( nd_wpd ) {
467: case 2:
468: if ( TD(d2) != TD(d1) ) return 0;
469: if ( d2[1] != d1[1] ) return 0;
470: return 1;
471: break;
472: case 3:
473: if ( TD(d2) != TD(d1) ) return 0;
474: if ( d2[1] != d1[1] ) return 0;
475: if ( d2[2] != d1[2] ) return 0;
476: return 1;
477: break;
478: default:
479: for ( i = 0; i < nd_wpd; i++ )
480: if ( *d1++ != *d2++ ) return 0;
481: return 1;
482: break;
483: }
1.1 noro 484: }
485:
1.61 noro 486: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 487: {
488: int i;
489:
490: switch ( nd_wpd ) {
1.41 noro 491: case 2:
1.34 noro 492: TD(d2) = TD(d1);
493: d2[1] = d1[1];
1.6 noro 494: break;
1.41 noro 495: case 3:
1.34 noro 496: TD(d2) = TD(d1);
1.6 noro 497: d2[1] = d1[1];
1.34 noro 498: d2[2] = d1[2];
1.6 noro 499: break;
500: default:
1.41 noro 501: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 502: d2[i] = d1[i];
503: break;
504: }
505: }
506:
1.61 noro 507: INLINE void ndl_zero(UINT *d)
508: {
509: int i;
510: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
511: }
512:
513: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 514: {
515: int i;
516:
1.43 noro 517: #if 1
1.6 noro 518: switch ( nd_wpd ) {
1.41 noro 519: case 2:
520: TD(d) = TD(d1)+TD(d2);
1.34 noro 521: d[1] = d1[1]+d2[1];
1.6 noro 522: break;
1.41 noro 523: case 3:
524: TD(d) = TD(d1)+TD(d2);
1.6 noro 525: d[1] = d1[1]+d2[1];
1.34 noro 526: d[2] = d1[2]+d2[2];
1.6 noro 527: break;
528: default:
1.43 noro 529: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 530: break;
531: }
1.43 noro 532: #else
533: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
534: #endif
1.6 noro 535: }
536:
1.55 noro 537: /* d1 += d2 */
1.61 noro 538: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 539: {
540: int i;
541:
542: #if 1
543: switch ( nd_wpd ) {
544: case 2:
545: TD(d1) += TD(d2);
546: d1[1] += d2[1];
547: break;
548: case 3:
549: TD(d1) += TD(d2);
550: d1[1] += d2[1];
551: d1[2] += d2[2];
552: break;
553: default:
554: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
555: break;
556: }
557: #else
558: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
559: #endif
560: }
561:
1.61 noro 562: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 563: {
564: int i;
565:
1.43 noro 566: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 567: }
568:
1.61 noro 569: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 570: {
1.61 noro 571: UINT t1,t2,u,u1,u2;
1.1 noro 572: int i,j;
573:
1.65 noro 574: #if USE_UNROLL
1.1 noro 575: switch ( nd_bpe ) {
1.62 noro 576: case 3:
577: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
578: u1 = d1[i]; u2 = d2[i];
579: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
580: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
581: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
582: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
583: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
584: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
585: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
586: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
587: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
588: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
589: }
590: return 1;
591: break;
1.1 noro 592: case 4:
1.41 noro 593: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 594: u1 = d1[i]; u2 = d2[i];
595: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 596: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
597: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
598: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
599: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
600: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
601: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
602: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 603: }
604: return 1;
605: break;
606: case 6:
1.41 noro 607: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 608: u1 = d1[i]; u2 = d2[i];
609: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 610: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
611: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
612: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
613: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 614: }
615: return 1;
616: break;
617: case 8:
1.41 noro 618: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 619: u1 = d1[i]; u2 = d2[i];
620: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 621: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
622: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
623: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 624: }
625: return 1;
626: break;
627: case 16:
1.41 noro 628: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 629: u1 = d1[i]; u2 = d2[i];
630: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 631: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 632: }
633: return 1;
634: break;
635: case 32:
1.41 noro 636: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 637: if ( d1[i] && d2[i] ) return 0;
638: return 1;
639: break;
640: default:
1.41 noro 641: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 642: u1 = d1[i]; u2 = d2[i];
643: for ( j = 0; j < nd_epw; j++ ) {
644: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
645: u1 >>= nd_bpe; u2 >>= nd_bpe;
646: }
647: }
648: return 1;
649: break;
650: }
1.65 noro 651: #else
652: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
653: u1 = d1[i]; u2 = d2[i];
654: for ( j = 0; j < nd_epw; j++ ) {
655: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
656: u1 >>= nd_bpe; u2 >>= nd_bpe;
657: }
658: }
659: return 1;
660: #endif
1.1 noro 661: }
662:
1.61 noro 663: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 664: {
1.61 noro 665: UINT u2;
666: UINT *d1;
1.5 noro 667: int i,j,ind,k;
1.1 noro 668:
1.5 noro 669: d1 = nd_bound[index];
670: ind = 0;
1.65 noro 671: #if USE_UNROLL
1.5 noro 672: switch ( nd_bpe ) {
1.62 noro 673: case 3:
674: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
675: u2 = d2[i];
676: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
677: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
678: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
679: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
680: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
681: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
682: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
683: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
684: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
685: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
686: }
687: return 0;
688: break;
1.5 noro 689: case 4:
1.41 noro 690: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 691: u2 = d2[i];
692: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
693: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
694: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
695: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
696: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
697: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
698: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
699: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
700: }
701: return 0;
702: break;
703: case 6:
1.41 noro 704: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 705: u2 = d2[i];
706: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
707: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
708: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
709: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
710: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
711: }
712: return 0;
713: break;
714: case 8:
1.41 noro 715: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 716: u2 = d2[i];
717: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
718: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
719: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
720: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
721: }
722: return 0;
723: break;
724: case 16:
1.41 noro 725: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 726: u2 = d2[i];
727: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
728: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
729: }
730: return 0;
731: break;
732: case 32:
1.41 noro 733: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 734: if ( d1[i]+d2[i]<d1[i] ) return 1;
735: return 0;
736: break;
737: default:
1.41 noro 738: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 739: u2 = d2[i];
740: k = (nd_epw-1)*nd_bpe;
741: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
742: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
743: }
744: return 0;
745: break;
746: }
1.65 noro 747: #else
748: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
749: u2 = d2[i];
750: k = (nd_epw-1)*nd_bpe;
751: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
752: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
753: }
754: return 0;
755: #endif
1.1 noro 756: }
757:
1.61 noro 758: INLINE int ndl_hash_value(UINT *d)
1.1 noro 759: {
760: int i;
761: int r;
762:
1.34 noro 763: r = 0;
1.41 noro 764: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 765: r = ((r<<16)+d[i])%REDTAB_LEN;
766: return r;
767: }
768:
1.63 noro 769: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 770: {
1.13 noro 771: RHist r;
1.6 noro 772: int d,k,i;
1.1 noro 773:
1.63 noro 774: d = ndl_hash_value(dg);
1.13 noro 775: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 776: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 777: if ( k > 0 ) nd_notfirst++;
778: nd_found++;
1.13 noro 779: return r->index;
1.1 noro 780: }
781: }
1.13 noro 782: if ( Reverse )
783: for ( i = nd_psn-1; i >= 0; i-- ) {
784: r = nd_psh[i];
1.34 noro 785: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 786: nd_create++;
1.34 noro 787: nd_append_red(dg,i);
1.13 noro 788: return i;
789: }
790: }
791: else
792: for ( i = 0; i < nd_psn; i++ ) {
793: r = nd_psh[i];
1.34 noro 794: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 795: nd_create++;
1.34 noro 796: nd_append_red(dg,i);
1.13 noro 797: return i;
798: }
1.1 noro 799: }
1.6 noro 800: return -1;
1.1 noro 801: }
802:
1.63 noro 803: ND nd_merge(ND p1,ND p2)
804: {
805: int n,c;
806: int t,can,td1,td2;
807: ND r;
808: NM m1,m2,mr0,mr,s;
809:
810: if ( !p1 ) return p2;
811: else if ( !p2 ) return p1;
812: else {
813: can = 0;
814: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
815: c = DL_COMPARE(DL(m1),DL(m2));
816: switch ( c ) {
817: case 0:
818: s = m1; m1 = NEXT(m1);
819: can++; NEXTNM2(mr0,mr,s);
820: s = m2; m2 = NEXT(m2); FREENM(s);
821: break;
822: case 1:
823: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
824: break;
825: case -1:
826: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
827: break;
828: }
829: }
830: if ( !mr0 )
831: if ( m1 ) mr0 = m1;
832: else if ( m2 ) mr0 = m2;
833: else return 0;
834: else if ( m1 ) NEXT(mr) = m1;
835: else if ( m2 ) NEXT(mr) = m2;
836: else NEXT(mr) = 0;
837: BDY(p1) = mr0;
838: SG(p1) = MAX(SG(p1),SG(p2));
839: LEN(p1) = LEN(p1)+LEN(p2)-can;
840: FREEND(p2);
841: return p1;
842: }
843: }
844:
1.31 noro 845: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 846: {
847: int n,c;
1.34 noro 848: int t,can,td1,td2;
1.1 noro 849: ND r;
850: NM m1,m2,mr0,mr,s;
851:
1.34 noro 852: if ( !p1 ) return p2;
853: else if ( !p2 ) return p1;
1.71 noro 854: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 855: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 856: else {
1.30 noro 857: can = 0;
1.1 noro 858: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 859: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 860: switch ( c ) {
861: case 0:
1.19 noro 862: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 863: if ( t < 0 ) t += mod;
1.1 noro 864: s = m1; m1 = NEXT(m1);
865: if ( t ) {
1.34 noro 866: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 867: } else {
1.34 noro 868: can += 2; FREENM(s);
1.17 noro 869: }
870: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 871: break;
872: case 1:
873: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
874: break;
875: case -1:
876: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
877: break;
878: }
879: }
880: if ( !mr0 )
881: if ( m1 ) mr0 = m1;
882: else if ( m2 ) mr0 = m2;
883: else return 0;
884: else if ( m1 ) NEXT(mr) = m1;
885: else if ( m2 ) NEXT(mr) = m2;
886: else NEXT(mr) = 0;
887: BDY(p1) = mr0;
888: SG(p1) = MAX(SG(p1),SG(p2));
889: LEN(p1) = LEN(p1)+LEN(p2)-can;
890: FREEND(p2);
891: return p1;
892: }
893: }
894:
895: /* XXX on opteron, the inlined manipulation of destructive additon of
896: * two NM seems to make gcc optimizer get confused, so the part is
897: * done in a function.
898: */
899:
900: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
901: {
902: NM s;
903: Q t;
904: int can;
905:
906: addq(CQ(*m1),CQ(*m2),&t);
907: s = *m1; *m1 = NEXT(*m1);
908: if ( t ) {
909: can = 1; NEXTNM2(*mr0,*mr,s); CQ(*mr) = (t);
910: } else {
911: can = 2; FREENM(s);
912: }
913: s = *m2; *m2 = NEXT(*m2); FREENM(s);
914: return can;
915: }
916:
917: ND nd_add_q(ND p1,ND p2)
918: {
919: int n,c,can;
920: ND r;
921: NM m1,m2,mr0,mr,s;
922: Q t;
923:
924: if ( !p1 ) return p2;
925: else if ( !p2 ) return p1;
926: else {
927: can = 0;
928: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
929: c = DL_COMPARE(DL(m1),DL(m2));
930: switch ( c ) {
931: case 0:
932: #if defined(__x86_64__)
933: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
934: #else
935: addq(CQ(m1),CQ(m2),&t);
936: s = m1; m1 = NEXT(m1);
937: if ( t ) {
938: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
939: } else {
940: can += 2; FREENM(s);
941: }
942: s = m2; m2 = NEXT(m2); FREENM(s);
943: #endif
1.17 noro 944: break;
945: case 1:
946: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
947: break;
948: case -1:
949: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
950: break;
951: }
952: }
953: if ( !mr0 )
1.34 noro 954: if ( m1 ) mr0 = m1;
955: else if ( m2 ) mr0 = m2;
956: else return 0;
957: else if ( m1 ) NEXT(mr) = m1;
958: else if ( m2 ) NEXT(mr) = m2;
959: else NEXT(mr) = 0;
1.17 noro 960: BDY(p1) = mr0;
961: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 962: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 963: FREEND(p2);
964: return p1;
965: }
966: }
967:
1.71 noro 968: ND nd_add_sf(ND p1,ND p2)
969: {
970: int n,c,can;
971: ND r;
972: NM m1,m2,mr0,mr,s;
973: int t;
974:
975: if ( !p1 ) return p2;
976: else if ( !p2 ) return p1;
977: else {
978: can = 0;
979: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
980: c = DL_COMPARE(DL(m1),DL(m2));
981: switch ( c ) {
982: case 0:
983: t = _addsf(CM(m1),CM(m2));
984: s = m1; m1 = NEXT(m1);
985: if ( t ) {
986: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
987: } else {
988: can += 2; FREENM(s);
989: }
990: s = m2; m2 = NEXT(m2); FREENM(s);
991: break;
992: case 1:
993: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
994: break;
995: case -1:
996: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
997: break;
998: }
999: }
1000: if ( !mr0 )
1001: if ( m1 ) mr0 = m1;
1002: else if ( m2 ) mr0 = m2;
1003: else return 0;
1004: else if ( m1 ) NEXT(mr) = m1;
1005: else if ( m2 ) NEXT(mr) = m2;
1006: else NEXT(mr) = 0;
1007: BDY(p1) = mr0;
1008: SG(p1) = MAX(SG(p1),SG(p2));
1009: LEN(p1) = LEN(p1)+LEN(p2)-can;
1010: FREEND(p2);
1011: return p1;
1012: }
1013: }
1014:
1.1 noro 1015: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1016: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1017: {
1.11 noro 1018: ND d;
1.1 noro 1019: NM m,mrd,tail;
1.7 noro 1020: NM mul;
1.10 noro 1021: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1022: int c,c1,c2,dummy;
1.17 noro 1023: RHist h;
1.11 noro 1024: NDV p,red;
1.69 noro 1025: Q cg,cred,gcd,tq,qq;
1.21 noro 1026: double hmag;
1.1 noro 1027:
1.69 noro 1028: if ( dn ) {
1029: if ( mod )
1030: dn->m = 1;
1031: else
1032: dn->z = ONE;
1033: }
1.1 noro 1034: if ( !g ) {
1035: *rp = 0;
1036: return 1;
1037: }
1.34 noro 1038: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1039:
1.14 noro 1040: sugar0 = sugar = SG(g);
1.1 noro 1041: n = NV(g);
1.61 noro 1042: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1043: for ( d = 0; g; ) {
1.63 noro 1044: index = ndl_find_reducer(HDL(g));
1.6 noro 1045: if ( index >= 0 ) {
1.17 noro 1046: h = nd_psh[index];
1047: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1048: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1049: nd_free(g); nd_free(d);
1050: return 0;
1051: }
1.77 noro 1052: if ( nd_demand )
1053: p = ndv_load(index);
1054: else
1055: p = ps[index];
1.71 noro 1056: if ( mod == -1 )
1057: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1058: else if ( mod ) {
1.19 noro 1059: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1060: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1061: } else {
1.17 noro 1062: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1063: chsgnq(cg,&CQ(mul));
1.20 noro 1064: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1065: if ( dn ) {
1066: mulq(dn->z,cred,&tq); dn->z = tq;
1067: }
1.16 noro 1068: }
1.55 noro 1069: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1070: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1071: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1072: tq = HCQ(g);
1.21 noro 1073: nd_removecont2(d,g);
1.69 noro 1074: if ( dn ) {
1075: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1076: }
1.21 noro 1077: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1078: }
1.1 noro 1079: } else if ( !full ) {
1080: *rp = g;
1081: return 1;
1082: } else {
1083: m = BDY(g);
1084: if ( NEXT(m) ) {
1.34 noro 1085: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1086: } else {
1087: FREEND(g); g = 0;
1088: }
1089: if ( d ) {
1.34 noro 1090: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1091: } else {
1.34 noro 1092: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1093: }
1094: }
1095: }
1.34 noro 1096: if ( d ) SG(d) = sugar;
1.1 noro 1097: *rp = d;
1098: return 1;
1099: }
1.28 noro 1100:
1.53 noro 1101: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1102: {
1103: int hindex,index;
1104: NDV p;
1105: ND u,d,red;
1106: NODE l;
1.31 noro 1107: NM mul,m,mrd,tail;
1.25 noro 1108: int sugar,psugar,n,h_reducible;
1109: PGeoBucket bucket;
1110: int c,c1,c2;
1.26 noro 1111: Q cg,cred,gcd,zzz;
1.25 noro 1112: RHist h;
1.28 noro 1113: double hmag,gmag;
1.77 noro 1114: int count = 0;
1115: int hcount = 0;
1.25 noro 1116:
1117: if ( !g ) {
1118: *rp = 0;
1119: return 1;
1120: }
1121: sugar = SG(g);
1122: n = NV(g);
1.34 noro 1123: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1124: bucket = create_pbucket();
1.31 noro 1125: add_pbucket(mod,bucket,g);
1.25 noro 1126: d = 0;
1.61 noro 1127: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1128: while ( 1 ) {
1.26 noro 1129: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1130: if ( hindex < 0 ) {
1.77 noro 1131: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1132: if ( d ) SG(d) = sugar;
1.25 noro 1133: *rp = d;
1134: return 1;
1135: }
1136: g = bucket->body[hindex];
1.63 noro 1137: index = ndl_find_reducer(HDL(g));
1.25 noro 1138: if ( index >= 0 ) {
1.77 noro 1139: count++;
1140: if ( !d ) hcount++;
1.25 noro 1141: h = nd_psh[index];
1142: ndl_sub(HDL(g),DL(h),DL(mul));
1143: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1144: nd_free(d);
1.25 noro 1145: free_pbucket(bucket);
1146: *rp = 0;
1147: return 0;
1148: }
1.53 noro 1149: p = ps[index];
1.71 noro 1150: if ( mod == -1 )
1151: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1152: else if ( mod ) {
1.25 noro 1153: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1154: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1155: } else {
1156: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1157: chsgnq(cg,&CQ(mul));
1.26 noro 1158: nd_mul_c_q(d,cred);
1159: mulq_pbucket(bucket,cred);
1160: g = bucket->body[hindex];
1.28 noro 1161: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1162: }
1.55 noro 1163: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1164: bucket->body[hindex] = nd_remove_head(g);
1165: red = nd_remove_head(red);
1.31 noro 1166: add_pbucket(mod,bucket,red);
1.34 noro 1167: psugar = SG(p)+TD(DL(mul));
1168: sugar = MAX(sugar,psugar);
1.28 noro 1169: if ( !mod && hmag && (gmag > hmag) ) {
1170: g = normalize_pbucket(mod,bucket);
1171: if ( !g ) {
1.34 noro 1172: if ( d ) SG(d) = sugar;
1.28 noro 1173: *rp = d;
1174: return 1;
1175: }
1176: nd_removecont2(d,g);
1177: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1178: add_pbucket(mod,bucket,g);
1.28 noro 1179: }
1.25 noro 1180: } else if ( !full ) {
1181: g = normalize_pbucket(mod,bucket);
1.34 noro 1182: if ( g ) SG(g) = sugar;
1.25 noro 1183: *rp = g;
1184: return 1;
1185: } else {
1186: m = BDY(g);
1187: if ( NEXT(m) ) {
1.34 noro 1188: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1189: } else {
1190: FREEND(g); g = 0;
1191: }
1192: bucket->body[hindex] = g;
1193: NEXT(m) = 0;
1194: if ( d ) {
1.34 noro 1195: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1196: } else {
1.34 noro 1197: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1198: }
1199: }
1200: }
1201: }
1.27 noro 1202:
1.61 noro 1203: /* input : list of NDV, cand : list of NDV */
1.28 noro 1204:
1.61 noro 1205: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1206: {
1207: int n,i,stat;
1208: ND nf,d;
1.61 noro 1209: NDV r;
1.45 noro 1210: NODE t,s;
1.69 noro 1211: union oNDC dn;
1.45 noro 1212:
1.61 noro 1213: ndv_setup(0,0,cand);
1.31 noro 1214: n = length(cand);
1.28 noro 1215:
1216: /* membercheck : list is a subset of Id(cand) ? */
1217: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1218: again:
1.61 noro 1219: if ( nd_bpe > obpe )
1220: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1221: else
1222: r = (NDV)BDY(t);
1223: d = ndvtond(0,r);
1.69 noro 1224: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1225: if ( !stat ) {
1226: nd_reconstruct(0,0,0);
1227: goto again;
1228: } else if ( nf ) return 0;
1.71 noro 1229: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1230: }
1.71 noro 1231: if ( DP_Print ) { printf("\n"); }
1.28 noro 1232: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1233: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1234: /* XXX */
1.23 noro 1235: return 1;
1236: }
1.1 noro 1237:
1238: ND nd_remove_head(ND p)
1239: {
1240: NM m;
1241:
1242: m = BDY(p);
1243: if ( !NEXT(m) ) {
1.34 noro 1244: FREEND(p); p = 0;
1.31 noro 1245: } else {
1.34 noro 1246: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1247: }
1.1 noro 1248: FREENM(m);
1249: return p;
1250: }
1251:
1.69 noro 1252: ND nd_separate_head(ND p,ND *head)
1253: {
1254: NM m,m0;
1255: ND r;
1256:
1257: m = BDY(p);
1258: if ( !NEXT(m) ) {
1259: *head = p; p = 0;
1260: } else {
1261: m0 = m;
1262: BDY(p) = NEXT(m); LEN(p)--;
1263: NEXT(m0) = 0;
1264: MKND(NV(p),m0,1,r);
1265: *head = r;
1266: }
1267: return p;
1268: }
1269:
1.1 noro 1270: PGeoBucket create_pbucket()
1271: {
1272: PGeoBucket g;
1273:
1274: g = CALLOC(1,sizeof(struct oPGeoBucket));
1275: g->m = -1;
1276: return g;
1277: }
1278:
1.25 noro 1279: void free_pbucket(PGeoBucket b) {
1280: int i;
1281:
1.26 noro 1282: for ( i = 0; i <= b->m; i++ )
1.25 noro 1283: if ( b->body[i] ) {
1284: nd_free(b->body[i]);
1285: b->body[i] = 0;
1286: }
1287: GC_free(b);
1288: }
1289:
1.63 noro 1290: void add_pbucket_symbolic(PGeoBucket g,ND d)
1291: {
1292: int l,i,k,m;
1293:
1294: if ( !d )
1295: return;
1296: l = LEN(d);
1297: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1298: /* 2^(k-1) < l <= 2^k (=m) */
1299: d = nd_merge(g->body[k],d);
1300: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1301: g->body[k] = 0;
1302: d = nd_merge(g->body[k+1],d);
1303: }
1304: g->body[k] = d;
1305: g->m = MAX(g->m,k);
1306: }
1307:
1.31 noro 1308: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1309: {
1.31 noro 1310: int l,i,k,m;
1.1 noro 1311:
1.31 noro 1312: if ( !d )
1313: return;
1314: l = LEN(d);
1.29 noro 1315: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1316: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1317: d = nd_add(mod,g->body[k],d);
1318: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1319: g->body[k] = 0;
1.31 noro 1320: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1321: }
1322: g->body[k] = d;
1323: g->m = MAX(g->m,k);
1324: }
1325:
1.26 noro 1326: void mulq_pbucket(PGeoBucket g,Q c)
1327: {
1328: int k;
1329:
1330: for ( k = 0; k <= g->m; k++ )
1331: nd_mul_c_q(g->body[k],c);
1332: }
1333:
1.63 noro 1334: NM remove_head_pbucket_symbolic(PGeoBucket g)
1335: {
1336: int j,i,k,c;
1337: NM head;
1338:
1339: k = g->m;
1340: j = -1;
1341: for ( i = 0; i <= k; i++ ) {
1342: if ( !g->body[i] ) continue;
1343: if ( j < 0 ) j = i;
1344: else {
1345: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1346: if ( c > 0 )
1347: j = i;
1348: else if ( c == 0 )
1349: g->body[i] = nd_remove_head(g->body[i]);
1350: }
1351: }
1352: if ( j < 0 ) return 0;
1353: else {
1354: head = BDY(g->body[j]);
1355: if ( !NEXT(head) ) {
1356: FREEND(g->body[j]);
1357: g->body[j] = 0;
1358: } else {
1359: BDY(g->body[j]) = NEXT(head);
1360: LEN(g->body[j])--;
1361: }
1362: return head;
1363: }
1364: }
1365:
1.19 noro 1366: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1367: {
1368: int j,i,c,k,nv,sum;
1.61 noro 1369: UINT *di,*dj;
1.1 noro 1370: ND gi,gj;
1371:
1372: k = g->m;
1373: while ( 1 ) {
1374: j = -1;
1375: for ( i = 0; i <= k; i++ ) {
1376: if ( !(gi = g->body[i]) )
1377: continue;
1378: if ( j < 0 ) {
1379: j = i;
1380: gj = g->body[j];
1381: dj = HDL(gj);
1.14 noro 1382: sum = HCM(gj);
1.1 noro 1383: } else {
1.34 noro 1384: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1385: if ( c > 0 ) {
1.34 noro 1386: if ( sum ) HCM(gj) = sum;
1387: else g->body[j] = nd_remove_head(gj);
1.1 noro 1388: j = i;
1389: gj = g->body[j];
1390: dj = HDL(gj);
1.14 noro 1391: sum = HCM(gj);
1.1 noro 1392: } else if ( c == 0 ) {
1.71 noro 1393: if ( mod == -1 )
1394: sum = _addsf(sum,HCM(gi));
1395: else {
1396: sum = sum+HCM(gi)-mod;
1397: if ( sum < 0 ) sum += mod;
1398: }
1.1 noro 1399: g->body[i] = nd_remove_head(gi);
1400: }
1401: }
1402: }
1.34 noro 1403: if ( j < 0 ) return -1;
1.1 noro 1404: else if ( sum ) {
1.14 noro 1405: HCM(gj) = sum;
1.26 noro 1406: return j;
1.31 noro 1407: } else
1.26 noro 1408: g->body[j] = nd_remove_head(gj);
1409: }
1410: }
1411:
1412: int head_pbucket_q(PGeoBucket g)
1413: {
1414: int j,i,c,k,nv;
1415: Q sum,t;
1416: ND gi,gj;
1417:
1418: k = g->m;
1419: while ( 1 ) {
1420: j = -1;
1421: for ( i = 0; i <= k; i++ ) {
1.34 noro 1422: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1423: if ( j < 0 ) {
1424: j = i;
1425: gj = g->body[j];
1426: sum = HCQ(gj);
1427: } else {
1428: nv = NV(gi);
1.34 noro 1429: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1430: if ( c > 0 ) {
1.34 noro 1431: if ( sum ) HCQ(gj) = sum;
1432: else g->body[j] = nd_remove_head(gj);
1.26 noro 1433: j = i;
1434: gj = g->body[j];
1435: sum = HCQ(gj);
1436: } else if ( c == 0 ) {
1437: addq(sum,HCQ(gi),&t);
1438: sum = t;
1439: g->body[i] = nd_remove_head(gi);
1440: }
1441: }
1442: }
1.34 noro 1443: if ( j < 0 ) return -1;
1.26 noro 1444: else if ( sum ) {
1445: HCQ(gj) = sum;
1.1 noro 1446: return j;
1.31 noro 1447: } else
1.1 noro 1448: g->body[j] = nd_remove_head(gj);
1449: }
1450: }
1451:
1.25 noro 1452: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1453: {
1.31 noro 1454: int i;
1.1 noro 1455: ND r,t;
1456:
1457: r = 0;
1.28 noro 1458: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1459: r = nd_add(mod,r,g->body[i]);
1.28 noro 1460: g->body[i] = 0;
1461: }
1462: g->m = -1;
1.1 noro 1463: return r;
1464: }
1465:
1.92 noro 1466: void do_diagonalize(int sugar,int m)
1467: {
1468: int i,nh,stat;
1469: NODE r,g,t;
1470: ND h,nf,s,head;
1471: NDV nfv;
1472: Q q,num,den;
1473: union oNDC dn;
1474:
1475: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1476: if ( nd_demand )
1477: nfv = ndv_load(i);
1478: else
1479: nfv = nd_ps[i];
1480: s = ndvtond(m,nfv);
1481: s = nd_separate_head(s,&head);
1482: nd_nf(m,s,nd_ps,1,&dn,&nf);
1483: if ( !m ) {
1484: NTOQ(NM(dn.z),SGN(dn.z),num);
1485: mulq(HCQ(head),num,&q); HCQ(head) = q;
1486: if ( DN(dn.z) ) {
1487: NTOQ(DN(dn.z),1,den);
1488: nd_mul_c_q(nf,den);
1489: }
1490: }
1491: nf = nd_add(m,head,nf);
1492: ndv_free(nfv);
1493: nd_removecont(m,nf);
1494: nfv = ndtondv(m,nf);
1495: nd_free(nf);
1496: nd_bound[i] = ndv_compute_bound(nfv);
1497: if ( nd_demand ) {
1498: ndv_save(nfv,i);
1499: ndv_free(nfv);
1500: } else
1501: nd_ps[i] = nfv;
1502: }
1503: }
1504:
1.27 noro 1505: /* return value = 0 => input is not a GB */
1506:
1.77 noro 1507: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1508: {
1509: int i,nh,sugar,stat;
1.23 noro 1510: NODE r,g,t;
1.1 noro 1511: ND_pairs d;
1512: ND_pairs l;
1.77 noro 1513: ND h,nf,s,head;
1.63 noro 1514: NDV nfv;
1.85 noro 1515: Q q,num,den;
1.69 noro 1516: union oNDC dn;
1.1 noro 1517:
1.23 noro 1518: g = 0; d = 0;
1519: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1520: d = update_pairs(d,g,i);
1521: g = update_base(g,i);
1522: }
1523: sugar = 0;
1524: while ( d ) {
1525: again:
1526: l = nd_minp(d,&d);
1.14 noro 1527: if ( SG(l) != sugar ) {
1.92 noro 1528: if ( ishomo ) do_diagonalize(sugar,m);
1529:
1.14 noro 1530: sugar = SG(l);
1.71 noro 1531: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1532: }
1.53 noro 1533: stat = nd_sp(m,0,l,&h);
1.1 noro 1534: if ( !stat ) {
1535: NEXT(l) = d; d = l;
1.20 noro 1536: d = nd_reconstruct(m,0,d);
1.1 noro 1537: goto again;
1538: }
1.41 noro 1539: #if USE_GEOBUCKET
1.69 noro 1540: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1541: #else
1.69 noro 1542: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1543: #endif
1.1 noro 1544: if ( !stat ) {
1545: NEXT(l) = d; d = l;
1.20 noro 1546: d = nd_reconstruct(m,0,d);
1.1 noro 1547: goto again;
1548: } else if ( nf ) {
1.27 noro 1549: if ( checkonly ) return 0;
1.71 noro 1550: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1551: nd_removecont(m,nf);
1552: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1553: nh = ndv_newps(m,nfv,0);
1.1 noro 1554: d = update_pairs(d,g,nh);
1555: g = update_base(g,nh);
1556: FREENDP(l);
1557: } else {
1.71 noro 1558: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1559: FREENDP(l);
1560: }
1561: }
1.77 noro 1562: if ( nd_demand )
1563: for ( t = g; t; t = NEXT(t) )
1564: BDY(t) = (pointer)ndv_load((int)BDY(t));
1565: else
1566: for ( t = g; t; t = NEXT(t) )
1567: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1568: return g;
1569: }
1570:
1.92 noro 1571: void do_diagonalize_trace(int sugar,int m)
1.91 noro 1572: {
1573: int i,nh,stat;
1574: NODE r,g,t;
1575: ND h,nf,nfq,s,head;
1576: NDV nfv,nfqv;
1577: Q q,den,num;
1578: union oNDC dn;
1579:
1580: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1581: /* for nd_ps */
1582: s = ndvtond(m,nd_ps[i]);
1583: s = nd_separate_head(s,&head);
1584: nd_nf_pbucket(m,s,nd_ps,1,&nf);
1585: nf = nd_add(m,head,nf);
1586: ndv_free(nd_ps[i]);
1587: nd_ps[i] = ndtondv(m,nf);
1588: nd_free(nf);
1589:
1590: /* for nd_ps_trace */
1591: if ( nd_demand )
1592: nfv = ndv_load(i);
1593: else
1594: nfv = nd_ps_trace[i];
1595: s = ndvtond(0,nfv);
1596: s = nd_separate_head(s,&head);
1597: nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1598: NTOQ(NM(dn.z),SGN(dn.z),num);
1599: mulq(HCQ(head),num,&q); HCQ(head) = q;
1600: if ( DN(dn.z) ) {
1601: NTOQ(DN(dn.z),1,den);
1602: nd_mul_c_q(nf,den);
1603: }
1604: nf = nd_add(0,head,nf);
1605: ndv_free(nfv);
1606: nd_removecont(0,nf);
1607: nfv = ndtondv(0,nf);
1608: nd_free(nf);
1609: nd_bound[i] = ndv_compute_bound(nfv);
1610: if ( nd_demand ) {
1611: ndv_save(nfv,i);
1612: ndv_free(nfv);
1613: } else
1614: nd_ps_trace[i] = nfv;
1615: }
1616: }
1617:
1.82 noro 1618: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1619: {
1620: int i,nh,sugar,stat;
1.23 noro 1621: NODE r,g,t;
1.20 noro 1622: ND_pairs d;
1623: ND_pairs l;
1.82 noro 1624: ND h,nf,nfq,s,head;
1.63 noro 1625: NDV nfv,nfqv;
1.85 noro 1626: Q q,den,num;
1.69 noro 1627: union oNDC dn;
1.20 noro 1628:
1.23 noro 1629: g = 0; d = 0;
1630: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1631: d = update_pairs(d,g,i);
1632: g = update_base(g,i);
1633: }
1634: sugar = 0;
1635: while ( d ) {
1636: again:
1637: l = nd_minp(d,&d);
1638: if ( SG(l) != sugar ) {
1.93 noro 1639: if ( ishomo ) do_diagonalize_trace(sugar,m);
1.20 noro 1640: sugar = SG(l);
1.71 noro 1641: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1642: }
1.53 noro 1643: stat = nd_sp(m,0,l,&h);
1.20 noro 1644: if ( !stat ) {
1645: NEXT(l) = d; d = l;
1646: d = nd_reconstruct(m,1,d);
1647: goto again;
1648: }
1.41 noro 1649: #if USE_GEOBUCKET
1.53 noro 1650: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1651: #else
1.69 noro 1652: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1653: #endif
1.20 noro 1654: if ( !stat ) {
1655: NEXT(l) = d; d = l;
1656: d = nd_reconstruct(m,1,d);
1657: goto again;
1658: } else if ( nf ) {
1.77 noro 1659: if ( nd_demand ) {
1660: nfqv = ndv_load(nd_psn);
1661: nfq = ndvtond(0,nfqv);
1662: } else
1663: nfq = 0;
1664: if ( !nfq ) {
1665: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1666: NEXT(l) = d; d = l;
1667: d = nd_reconstruct(m,1,d);
1668: goto again;
1669: }
1.70 noro 1670: }
1.20 noro 1671: if ( nfq ) {
1.70 noro 1672: /* m|HC(nfq) => failure */
1.63 noro 1673: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1674:
1.71 noro 1675: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1676: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1677: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 noro 1678: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1679: d = update_pairs(d,g,nh);
1680: g = update_base(g,nh);
1681: } else {
1.71 noro 1682: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1683: }
1684: } else {
1.71 noro 1685: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1686: }
1687: FREENDP(l);
1688: }
1.77 noro 1689: if ( nd_demand )
1690: for ( t = g; t; t = NEXT(t) )
1691: BDY(t) = (pointer)ndv_load((int)BDY(t));
1692: else
1693: for ( t = g; t; t = NEXT(t) )
1694: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1695: return g;
1696: }
1697:
1.23 noro 1698: int ndv_compare(NDV *p1,NDV *p2)
1699: {
1.34 noro 1700: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1701: }
1702:
1703: int ndv_compare_rev(NDV *p1,NDV *p2)
1704: {
1.34 noro 1705: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1706: }
1707:
1.61 noro 1708: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1709: {
1.69 noro 1710: int i,n,stat;
1711: ND nf,g,head;
1.23 noro 1712: NODE t,a0,a;
1.69 noro 1713: union oNDC dn;
1714: NDV *w;
1.85 noro 1715: Q q,num,den;
1.23 noro 1716:
1.69 noro 1717: n = length(f);
1718: #if 0
1719: w = (NDV *)ALLOCA(n*sizeof(NDV));
1720: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1721: qsort(w,n,sizeof(NDV),
1722: (int (*)(const void *,const void *))ndv_compare);
1723: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1724: #endif
1725: ndv_setup(m,0,f);
1726: for ( i = 0; i < n; ) {
1727: g = ndvtond(m,nd_ps[i]);
1728: g = nd_separate_head(g,&head);
1729: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1730: if ( !stat )
1.69 noro 1731: nd_reconstruct(m,0,0);
1732: else {
1.71 noro 1733: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1734: if ( !m ) {
1735: NTOQ(NM(dn.z),SGN(dn.z),num);
1736: mulq(HCQ(head),num,&q); HCQ(head) = q;
1737: if ( DN(dn.z) ) {
1738: NTOQ(DN(dn.z),1,den);
1739: nd_mul_c_q(nf,den);
1740: }
1741: }
1.69 noro 1742: nf = nd_add(m,head,nf);
1743: ndv_free(nd_ps[i]);
1.24 noro 1744: nd_removecont(m,nf);
1.69 noro 1745: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1746: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1747: i++;
1.23 noro 1748: }
1749: }
1.71 noro 1750: if ( DP_Print ) { printf("\n"); }
1.23 noro 1751: for ( a0 = 0, i = 0; i < n; i++ ) {
1752: NEXTNODE(a0,a);
1.69 noro 1753: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1754: }
1755: NEXT(a) = 0;
1756: return a0;
1757: }
1758:
1.1 noro 1759: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1760: {
1761: ND_pairs d1,nd,cur,head,prev,remove;
1762:
1763: if ( !g ) return d;
1764: d = crit_B(d,t);
1765: d1 = nd_newpairs(g,t);
1766: d1 = crit_M(d1);
1767: d1 = crit_F(d1);
1.55 noro 1768: if ( do_weyl )
1769: head = d1;
1770: else {
1771: prev = 0; cur = head = d1;
1772: while ( cur ) {
1773: if ( crit_2( cur->i1,cur->i2 ) ) {
1774: remove = cur;
1775: if ( !prev ) head = cur = NEXT(cur);
1776: else cur = NEXT(prev) = NEXT(cur);
1777: FREENDP(remove);
1778: } else {
1779: prev = cur; cur = NEXT(cur);
1780: }
1.1 noro 1781: }
1782: }
1783: if ( !d )
1784: return head;
1785: else {
1786: nd = d;
1.34 noro 1787: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1788: NEXT(nd) = head;
1789: return d;
1790: }
1791: }
1792:
1793: ND_pairs nd_newpairs( NODE g, int t )
1794: {
1795: NODE h;
1.61 noro 1796: UINT *dl;
1.34 noro 1797: int ts,s;
1.1 noro 1798: ND_pairs r,r0;
1799:
1.20 noro 1800: dl = DL(nd_psh[t]);
1.34 noro 1801: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1802: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1803: NEXTND_pairs(r0,r);
1804: r->i1 = (int)BDY(h);
1805: r->i2 = t;
1.20 noro 1806: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1807: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1808: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1809: }
1810: NEXT(r) = 0;
1811: return r0;
1812: }
1813:
1814: ND_pairs crit_B( ND_pairs d, int s )
1815: {
1816: ND_pairs cur,head,prev,remove;
1.61 noro 1817: UINT *t,*tl,*lcm;
1.1 noro 1818: int td,tdl;
1819:
1820: if ( !d ) return 0;
1.20 noro 1821: t = DL(nd_psh[s]);
1.1 noro 1822: prev = 0;
1823: head = cur = d;
1.61 noro 1824: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1825: while ( cur ) {
1826: tl = cur->lcm;
1827: if ( ndl_reducible(tl,t)
1.20 noro 1828: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1829: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1830: remove = cur;
1831: if ( !prev ) {
1832: head = cur = NEXT(cur);
1833: } else {
1834: cur = NEXT(prev) = NEXT(cur);
1835: }
1836: FREENDP(remove);
1837: } else {
1.34 noro 1838: prev = cur; cur = NEXT(cur);
1.1 noro 1839: }
1840: }
1841: return head;
1842: }
1843:
1844: ND_pairs crit_M( ND_pairs d1 )
1845: {
1846: ND_pairs e,d2,d3,dd,p;
1.61 noro 1847: UINT *id,*jd;
1.1 noro 1848:
1849: for ( dd = 0, e = d1; e; e = d3 ) {
1850: if ( !(d2 = NEXT(e)) ) {
1851: NEXT(e) = dd;
1852: return e;
1853: }
1.34 noro 1854: id = LCM(e);
1.1 noro 1855: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1856: p = NEXT(d2);
1857: jd = LCM(d2);
1858: if ( ndl_equal(jd,id) )
1859: ;
1860: else if ( TD(jd) > TD(id) )
1.1 noro 1861: if ( ndl_reducible(jd,id) ) continue;
1862: else ;
1.34 noro 1863: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1864: NEXT(d2) = d3;
1865: d3 = d2;
1866: }
1867: NEXT(e) = dd;
1868: dd = e;
1869: continue;
1870: /**/
1871: delit: NEXT(d2) = d3;
1872: d3 = d2;
1873: for ( ; p; p = d2 ) {
1874: d2 = NEXT(p);
1875: NEXT(p) = d3;
1876: d3 = p;
1877: }
1878: FREENDP(e);
1879: }
1880: return dd;
1881: }
1882:
1883: ND_pairs crit_F( ND_pairs d1 )
1884: {
1885: ND_pairs rest, head,remove;
1886: ND_pairs last, p, r, w;
1887: int s;
1888:
1889: for ( head = last = 0, p = d1; NEXT(p); ) {
1890: r = w = equivalent_pairs(p,&rest);
1.14 noro 1891: s = SG(r);
1.1 noro 1892: w = NEXT(w);
1893: while ( w ) {
1894: if ( crit_2(w->i1,w->i2) ) {
1895: r = w;
1896: w = NEXT(w);
1897: while ( w ) {
1898: remove = w;
1899: w = NEXT(w);
1900: FREENDP(remove);
1901: }
1902: break;
1.14 noro 1903: } else if ( SG(w) < s ) {
1.1 noro 1904: FREENDP(r);
1905: r = w;
1.14 noro 1906: s = SG(r);
1.1 noro 1907: w = NEXT(w);
1908: } else {
1909: remove = w;
1910: w = NEXT(w);
1911: FREENDP(remove);
1912: }
1913: }
1914: if ( last ) NEXT(last) = r;
1915: else head = r;
1916: NEXT(last = r) = 0;
1917: p = rest;
1918: if ( !p ) return head;
1919: }
1920: if ( !last ) return p;
1921: NEXT(last) = p;
1922: return head;
1923: }
1924:
1925: int crit_2( int dp1, int dp2 )
1926: {
1.20 noro 1927: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 1928: }
1929:
1.40 noro 1930: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 1931: {
1932: ND_pairs w,p,r,s;
1.61 noro 1933: UINT *d;
1.1 noro 1934:
1935: w = d1;
1.34 noro 1936: d = LCM(w);
1.1 noro 1937: s = NEXT(w);
1938: NEXT(w) = 0;
1939: for ( r = 0; s; s = p ) {
1940: p = NEXT(s);
1.34 noro 1941: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 1942: NEXT(s) = w; w = s;
1.1 noro 1943: } else {
1.39 noro 1944: NEXT(s) = r; r = s;
1.1 noro 1945: }
1946: }
1947: *prest = r;
1948: return w;
1949: }
1950:
1951: NODE update_base(NODE nd,int ndp)
1952: {
1.61 noro 1953: UINT *dl, *dln;
1.1 noro 1954: NODE last, p, head;
1955:
1.20 noro 1956: dl = DL(nd_psh[ndp]);
1.1 noro 1957: for ( head = last = 0, p = nd; p; ) {
1.20 noro 1958: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 1959: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 1960: p = NEXT(p);
1961: if ( last ) NEXT(last) = p;
1962: } else {
1963: if ( !last ) head = p;
1964: p = NEXT(last = p);
1965: }
1966: }
1967: head = append_one(head,ndp);
1968: return head;
1969: }
1970:
1971: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
1972: {
1973: ND_pairs m,ml,p,l;
1.61 noro 1974: UINT *lcm;
1.33 noro 1975: int s,td,len,tlen,c,c1;
1.1 noro 1976:
1977: if ( !(p = NEXT(m = d)) ) {
1978: *prest = p;
1979: NEXT(m) = 0;
1980: return m;
1981: }
1.14 noro 1982: s = SG(m);
1.77 noro 1983: if ( !NoSugar ) {
1984: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1985: if ( (SG(p) < s)
1986: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
1987: ml = l; m = p; s = SG(m);
1988: }
1989: } else {
1990: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1991: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
1992: ml = l; m = p; s = SG(m);
1993: }
1994: }
1.1 noro 1995: if ( !ml ) *prest = NEXT(m);
1996: else {
1997: NEXT(ml) = NEXT(m);
1998: *prest = d;
1999: }
2000: NEXT(m) = 0;
2001: return m;
2002: }
2003:
1.63 noro 2004: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2005: {
1.69 noro 2006: int msugar,i;
1.63 noro 2007: ND_pairs t,dm0,dm,dr0,dr;
2008:
2009: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2010: if ( SG(t) < msugar ) msugar = SG(t);
2011: dm0 = 0; dr0 = 0;
1.69 noro 2012: for ( i = 0, t = d; t; t = NEXT(t) )
2013: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2014: if ( dm0 ) NEXT(dm) = t;
2015: else dm0 = t;
2016: dm = t;
1.69 noro 2017: i++;
1.63 noro 2018: } else {
2019: if ( dr0 ) NEXT(dr) = t;
2020: else dr0 = t;
2021: dr = t;
2022: }
2023: NEXT(dm) = 0;
2024: if ( dr0 ) NEXT(dr) = 0;
2025: *prest = dr0;
2026: return dm0;
2027: }
2028:
1.77 noro 2029: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2030: {
1.3 noro 2031: int len;
1.13 noro 2032: RHist r;
1.20 noro 2033: NDV b;
1.3 noro 2034:
1.1 noro 2035: if ( nd_psn == nd_pslen ) {
2036: nd_pslen *= 2;
1.11 noro 2037: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2038: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2039: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2040: nd_bound = (UINT **)
2041: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2042: }
1.39 noro 2043: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2044: nd_ps[nd_psn] = a;
1.39 noro 2045: if ( aq ) {
1.63 noro 2046: nd_ps_trace[nd_psn] = aq;
2047: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2048: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2049: } else {
1.63 noro 2050: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2051: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2052: }
1.77 noro 2053: if ( nd_demand ) {
2054: if ( aq ) {
2055: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2056: nd_ps_trace[nd_psn] = 0;
2057: } else {
2058: ndv_save(nd_ps[nd_psn],nd_psn);
2059: nd_ps[nd_psn] = 0;
2060: }
2061: }
1.1 noro 2062: return nd_psn++;
2063: }
2064:
1.61 noro 2065: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2066: {
1.5 noro 2067: int i,j,td,len,max;
1.1 noro 2068: NODE s,s0,f0;
1.61 noro 2069: UINT *d;
1.13 noro 2070: RHist r;
1.69 noro 2071: NDV *w;
1.61 noro 2072: NDV a,am;
1.11 noro 2073:
2074: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2075:
1.69 noro 2076: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2077: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2078: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2079: qsort(w,nd_psn,sizeof(NDV),
2080: (int (*)(const void *,const void *))ndv_compare);
2081: nd_pslen = 2*nd_psn;
1.11 noro 2082: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2083: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2084: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2085: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2086:
1.11 noro 2087: if ( !nd_red )
1.13 noro 2088: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2089: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2090: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2091: if ( trace ) {
1.69 noro 2092: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2093: ndv_removecont(0,a);
2094: am = nd_ps[i] = ndv_dup(mod,a);
2095: ndv_mod(mod,am);
2096: ndv_removecont(mod,am);
2097: } else {
1.69 noro 2098: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2099: ndv_removecont(mod,a);
1.39 noro 2100: }
1.61 noro 2101: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2102: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2103: nd_psh[i] = r;
1.77 noro 2104: if ( nd_demand ) {
2105: if ( trace ) {
2106: ndv_save(nd_ps_trace[i],i);
2107: nd_ps_trace[i] = 0;
2108: } else {
2109: ndv_save(nd_ps[i],i);
2110: nd_ps[i] = 0;
2111: }
2112: }
1.20 noro 2113: }
2114: }
2115:
1.63 noro 2116: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2117: {
1.61 noro 2118: VL tv,fv,vv,vc;
1.1 noro 2119: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2120: int e,max,nvar;
1.69 noro 2121: NDV b;
1.77 noro 2122: int ishomo;
1.1 noro 2123:
1.78 noro 2124: if ( !m && Demand ) nd_demand = 1;
2125: else nd_demand = 0;
2126:
1.74 noro 2127: ndv_alloc = 0;
1.1 noro 2128: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2129: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2130: switch ( ord->id ) {
2131: case 1:
2132: if ( ord->nv != nvar )
2133: error("nd_{gr,f4} : invalid order specification");
2134: break;
2135: default:
2136: break;
2137: }
1.32 noro 2138: nd_init_ord(ord);
1.61 noro 2139: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2140: for ( tv = vv; tv; tv = NEXT(tv) ) {
2141: e = getdeg(tv->v,(P)BDY(t));
2142: max = MAX(e,max);
2143: }
2144: nd_setup_parameters(nvar,max);
1.77 noro 2145: ishomo = 1;
1.1 noro 2146: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2147: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2148: if ( ishomo )
2149: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2150: if ( m ) ndv_mod(m,b);
1.61 noro 2151: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2152: }
2153: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2154: ndv_setup(m,0,fd0);
1.77 noro 2155: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2156: nd_demand = 0;
1.61 noro 2157: x = ndv_reducebase(x);
2158: x = ndv_reduceall(m,x);
1.23 noro 2159: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2160: NEXTNODE(r0,r);
1.61 noro 2161: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2162: }
2163: if ( r0 ) NEXT(r) = 0;
2164: MKLIST(*rp,r0);
1.74 noro 2165: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2166: }
2167:
1.52 noro 2168: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2169: {
1.89 noro 2170: struct order_spec *ord1;
1.61 noro 2171: VL tv,fv,vv,vc;
1.27 noro 2172: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2173: int m,nocheck,nvar,mindex,e,max;
2174: NDV c;
2175: NMV a;
1.27 noro 2176: P p;
1.61 noro 2177: EPOS oepos;
1.82 noro 2178: int obpe,oadv,wmax,i,len,cbpe,ishomo;
1.20 noro 2179:
2180: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2181: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2182: switch ( ord->id ) {
2183: case 1:
2184: if ( ord->nv != nvar )
2185: error("nd_gr_trace : invalid order specification");
2186: break;
2187: default:
2188: break;
2189: }
1.52 noro 2190: nocheck = 0;
2191: mindex = 0;
2192:
1.78 noro 2193: if ( Demand ) nd_demand = 1;
2194: else nd_demand = 0;
2195:
1.52 noro 2196: /* setup modulus */
2197: if ( trace < 0 ) {
2198: trace = -trace;
2199: nocheck = 1;
2200: }
2201: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2202: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2203: for ( tv = vv; tv; tv = NEXT(tv) ) {
2204: e = getdeg(tv->v,(P)BDY(t));
2205: max = MAX(e,max);
1.23 noro 2206: }
1.61 noro 2207: nd_init_ord(ord);
2208: nd_setup_parameters(nvar,max);
2209: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2210: ishomo = 1;
1.61 noro 2211: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2212: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2213: if ( ishomo )
2214: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2215: if ( c ) {
2216: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2217: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2218: }
1.61 noro 2219: }
2220: if ( in0 ) NEXT(in) = 0;
2221: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2222: if ( !ishomo && homo ) {
1.61 noro 2223: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2224: c = (NDV)BDY(t); len = LEN(c);
2225: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2226: wmax = MAX(TD(DL(a)),wmax);
2227: }
2228: homogenize_order(ord,nvar,&ord1);
1.89 noro 2229: nd_init_ord(ord1);
1.61 noro 2230: nd_setup_parameters(nvar+1,wmax);
2231: for ( t = fd0; t; t = NEXT(t) )
2232: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2233: }
1.52 noro 2234: while ( 1 ) {
1.77 noro 2235: if ( Demand )
2236: nd_demand = 1;
1.61 noro 2237: ndv_setup(m,1,fd0);
1.82 noro 2238: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2239: if ( !cand ) {
2240: /* failure */
1.61 noro 2241: if ( trace > 1 ) { *rp = 0; return; }
2242: else m = get_lprime(++mindex);
1.52 noro 2243: continue;
2244: }
1.82 noro 2245: if ( !ishomo && homo ) {
1.27 noro 2246: /* dehomogenization */
1.61 noro 2247: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2248: nd_init_ord(ord);
1.61 noro 2249: nd_setup_parameters(nvar,0);
1.27 noro 2250: }
1.77 noro 2251: nd_demand = 0;
1.61 noro 2252: cand = ndv_reducebase(cand);
2253: cand = ndv_reduceall(0,cand);
1.78 noro 2254: cbpe = nd_bpe;
1.61 noro 2255: if ( nocheck )
2256: break;
2257: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2258: /* success */
2259: break;
2260: else if ( trace > 1 ) {
2261: /* failure */
2262: *rp = 0; return;
1.61 noro 2263: } else {
1.52 noro 2264: /* try the next modulus */
2265: m = get_lprime(++mindex);
1.61 noro 2266: /* reset the parameters */
1.82 noro 2267: if ( !ishomo && homo ) {
1.89 noro 2268: nd_init_ord(ord1);
1.61 noro 2269: nd_setup_parameters(nvar+1,wmax);
2270: } else {
2271: nd_init_ord(ord);
2272: nd_setup_parameters(nvar,max);
2273: }
2274: }
1.52 noro 2275: }
1.27 noro 2276: /* dp->p */
1.61 noro 2277: nd_bpe = cbpe;
1.73 noro 2278: nd_setup_parameters(nd_nvar,0);
1.61 noro 2279: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2280: MKLIST(*rp,cand);
1.1 noro 2281: }
2282:
1.61 noro 2283: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2284: {
1.61 noro 2285: UINT *d;
1.57 noro 2286: int i,j,l,s,ord_l;
1.43 noro 2287: struct order_pair *op;
1.1 noro 2288:
2289: d = dl->d;
1.41 noro 2290: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2291: if ( nd_blockmask ) {
2292: l = nd_blockmask->n;
2293: op = nd_blockmask->order_pair;
2294: for ( j = 0, s = 0; j < l; j++ ) {
2295: ord_l = op[j].length;
1.57 noro 2296: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2297: }
2298: TD(r) = ndl_weight(r);
1.61 noro 2299: ndl_weight_mask(r);
1.43 noro 2300: } else {
1.56 noro 2301: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2302: TD(r) = ndl_weight(r);
2303: }
1.1 noro 2304: }
2305:
1.61 noro 2306: DL ndltodl(int n,UINT *ndl)
1.1 noro 2307: {
2308: DL dl;
2309: int *d;
1.57 noro 2310: int i,j,l,s,ord_l;
1.43 noro 2311: struct order_pair *op;
1.1 noro 2312:
2313: NEWDL(dl,n);
1.34 noro 2314: dl->td = TD(ndl);
1.1 noro 2315: d = dl->d;
1.43 noro 2316: if ( nd_blockmask ) {
2317: l = nd_blockmask->n;
2318: op = nd_blockmask->order_pair;
2319: for ( j = 0, s = 0; j < l; j++ ) {
2320: ord_l = op[j].length;
1.57 noro 2321: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2322: }
2323: } else {
1.56 noro 2324: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2325: }
1.1 noro 2326: return dl;
2327: }
2328:
1.61 noro 2329: void ndl_print(UINT *dl)
1.1 noro 2330: {
2331: int n;
1.57 noro 2332: int i,j,l,ord_l,s,s0;
1.43 noro 2333: struct order_pair *op;
1.1 noro 2334:
2335: n = nd_nvar;
2336: printf("<<");
1.43 noro 2337: if ( nd_blockmask ) {
2338: l = nd_blockmask->n;
2339: op = nd_blockmask->order_pair;
2340: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2341: ord_l = op[j].length;
1.57 noro 2342: for ( i = 0; i < ord_l; i++, s++ )
2343: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2344: }
2345: } else {
1.56 noro 2346: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2347: }
1.1 noro 2348: printf(">>");
2349: }
2350:
2351: void nd_print(ND p)
2352: {
2353: NM m;
2354:
2355: if ( !p )
2356: printf("0\n");
2357: else {
2358: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2359: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2360: else printf("+%d*",CM(m));
1.14 noro 2361: ndl_print(DL(m));
1.1 noro 2362: }
2363: printf("\n");
2364: }
2365: }
2366:
1.16 noro 2367: void nd_print_q(ND p)
2368: {
2369: NM m;
2370:
2371: if ( !p )
2372: printf("0\n");
2373: else {
2374: for ( m = BDY(p); m; m = NEXT(m) ) {
2375: printf("+");
1.74 noro 2376: printexpr(CO,(Obj)CQ(m));
1.16 noro 2377: printf("*");
2378: ndl_print(DL(m));
2379: }
2380: printf("\n");
2381: }
2382: }
2383:
1.1 noro 2384: void ndp_print(ND_pairs d)
2385: {
2386: ND_pairs t;
2387:
1.34 noro 2388: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2389: printf("\n");
2390: }
2391:
1.20 noro 2392: void nd_removecont(int mod,ND p)
1.16 noro 2393: {
2394: int i,n;
2395: Q *w;
2396: Q dvr,t;
2397: NM m;
1.21 noro 2398: struct oVECT v;
2399: N q,r;
1.16 noro 2400:
1.71 noro 2401: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2402: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2403: else {
2404: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2405: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2406: v.len = n;
2407: v.body = (pointer *)w;
1.34 noro 2408: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2409: removecont_array(w,n);
2410: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2411: }
2412: }
2413:
1.21 noro 2414: void nd_removecont2(ND p1,ND p2)
2415: {
2416: int i,n1,n2,n;
2417: Q *w;
2418: Q dvr,t;
2419: NM m;
2420: struct oVECT v;
2421: N q,r;
2422:
2423: if ( !p1 ) {
2424: nd_removecont(0,p2); return;
2425: } else if ( !p2 ) {
2426: nd_removecont(0,p1); return;
2427: }
2428: n1 = nd_length(p1);
2429: n2 = nd_length(p2);
2430: n = n1+n2;
2431: w = (Q *)ALLOCA(n*sizeof(Q));
2432: v.len = n;
2433: v.body = (pointer *)w;
1.34 noro 2434: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2435: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2436: removecont_array(w,n);
2437: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2438: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2439: }
2440:
1.20 noro 2441: void ndv_removecont(int mod,NDV p)
1.16 noro 2442: {
2443: int i,len;
2444: Q *w;
2445: Q dvr,t;
2446: NMV m;
2447:
1.71 noro 2448: if ( mod == -1 )
2449: ndv_mul_c(mod,p,_invsf(HCM(p)));
2450: else if ( mod )
1.20 noro 2451: ndv_mul_c(mod,p,invm(HCM(p),mod));
2452: else {
2453: len = p->len;
2454: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2455: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2456: sortbynm(w,len);
2457: qltozl(w,len,&dvr);
2458: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2459: divq(CQ(m),dvr,&t); CQ(m) = t;
2460: }
1.16 noro 2461: }
1.21 noro 2462: }
2463:
1.61 noro 2464: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2465: {
2466: int len,i,max;
2467: NMV m,mr0,mr,t;
2468:
2469: len = p->len;
2470: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2471: max = MAX(max,TD(DL(m)));
2472: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2473: m = (NMV)((char *)mr0+(len-1)*oadv);
2474: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2475: t = (NMV)ALLOCA(nmv_adv);
2476: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2477: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2478: CQ(mr) = CQ(m);
2479: ndl_copy(DL(t),DL(mr));
2480: }
2481: NV(p)++;
2482: BDY(p) = mr0;
2483: }
2484:
1.45 noro 2485: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2486: {
1.45 noro 2487: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2488: Q *w;
2489: Q dvr,t;
2490: NMV m,r;
2491:
2492: len = p->len;
2493: newnvar = nd_nvar-1;
1.48 noro 2494: newexporigin = nd_get_exporigin(ord);
1.45 noro 2495: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2496: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2497: ndl_dehomogenize(DL(m));
1.23 noro 2498: if ( newwpd != nd_wpd ) {
1.90 noro 2499: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2500: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2501: CQ(r) = CQ(m);
2502: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2503: adj = nd_exporigin-newexporigin;
2504: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2505: }
2506: }
2507: NV(p)--;
2508: }
2509:
1.21 noro 2510: void removecont_array(Q *c,int n)
2511: {
2512: struct oVECT v;
2513: Q d0,d1,a,u,u1,gcd;
2514: int i;
2515: N qn,rn,gn;
2516: Q *q,*r;
2517:
2518: q = (Q *)ALLOCA(n*sizeof(Q));
2519: r = (Q *)ALLOCA(n*sizeof(Q));
2520: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2521: igcdv_estimate(&v,&d0);
2522: for ( i = 0; i < n; i++ ) {
2523: divn(NM(c[i]),NM(d0),&qn,&rn);
2524: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2525: NTOQ(rn,SGN(c[i]),r[i]);
2526: }
1.34 noro 2527: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2528: if ( i < n ) {
2529: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2530: igcdv(&v,&d1);
2531: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2532: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2533: for ( i = 0; i < n; i++ ) {
2534: mulq(a,q[i],&u);
2535: if ( r[i] ) {
2536: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2537: addq(u,u1,&q[i]);
2538: } else
2539: q[i] = u;
2540: }
2541: }
1.34 noro 2542: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2543: }
2544:
1.19 noro 2545: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2546: {
2547: NM m;
2548: int c,c1;
2549:
1.34 noro 2550: if ( !p ) return;
1.71 noro 2551: if ( mod == -1 )
2552: for ( m = BDY(p); m; m = NEXT(m) )
2553: CM(m) = _mulsf(CM(m),mul);
2554: else
2555: for ( m = BDY(p); m; m = NEXT(m) ) {
2556: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2557: }
1.1 noro 2558: }
2559:
1.16 noro 2560: void nd_mul_c_q(ND p,Q mul)
2561: {
2562: NM m;
2563: Q c;
2564:
1.34 noro 2565: if ( !p ) return;
1.16 noro 2566: for ( m = BDY(p); m; m = NEXT(m) ) {
2567: mulq(CQ(m),mul,&c); CQ(m) = c;
2568: }
2569: }
2570:
1.61 noro 2571: void nd_mul_c_p(VL vl,ND p,P mul)
2572: {
2573: NM m;
2574: P c;
2575:
2576: if ( !p ) return;
2577: for ( m = BDY(p); m; m = NEXT(m) ) {
2578: mulp(vl,CP(m),mul,&c); CP(m) = c;
2579: }
2580: }
2581:
1.1 noro 2582: void nd_free(ND p)
2583: {
2584: NM t,s;
2585:
1.34 noro 2586: if ( !p ) return;
1.1 noro 2587: t = BDY(p);
2588: while ( t ) {
2589: s = NEXT(t);
2590: FREENM(t);
2591: t = s;
2592: }
2593: FREEND(p);
2594: }
2595:
1.23 noro 2596: void ndv_free(NDV p)
2597: {
2598: GC_free(BDY(p));
2599: }
2600:
1.61 noro 2601: void nd_append_red(UINT *d,int i)
1.1 noro 2602: {
1.13 noro 2603: RHist m,m0;
1.1 noro 2604: int h;
2605:
1.13 noro 2606: NEWRHist(m);
1.34 noro 2607: h = ndl_hash_value(d);
1.13 noro 2608: m->index = i;
1.14 noro 2609: ndl_copy(d,DL(m));
1.1 noro 2610: NEXT(m) = nd_red[h];
2611: nd_red[h] = m;
2612: }
2613:
1.61 noro 2614: UINT *ndv_compute_bound(NDV p)
1.1 noro 2615: {
1.61 noro 2616: UINT *d1,*d2,*t;
2617: UINT u;
1.57 noro 2618: int i,j,k,l,len,ind;
1.45 noro 2619: NMV m;
1.1 noro 2620:
2621: if ( !p )
2622: return 0;
1.61 noro 2623: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2624: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2625: len = LEN(p);
2626: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2627: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2628: ndl_lcm(DL(m),d1,d2);
1.1 noro 2629: t = d1; d1 = d2; d2 = t;
2630: }
1.12 noro 2631: l = nd_nvar+31;
1.61 noro 2632: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2633: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2634: u = d1[i];
2635: k = (nd_epw-1)*nd_bpe;
2636: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2637: t[ind] = (u>>k)&nd_mask0;
2638: }
2639: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2640: return t;
2641: }
2642:
1.48 noro 2643: int nd_get_exporigin(struct order_spec *ord)
2644: {
1.51 noro 2645: switch ( ord->id ) {
1.96 ! noro 2646: case 0: case 2:
1.48 noro 2647: return 1;
1.41 noro 2648: case 1:
2649: /* block order */
1.43 noro 2650: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2651: return ord->ord.block.length+1;
1.96 ! noro 2652: case 3:
! 2653: error("nd_get_exporigin : composite order is not supported yet.");
1.41 noro 2654: }
1.48 noro 2655: }
2656:
1.61 noro 2657: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2658: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2659: struct order_pair *op;
1.48 noro 2660:
1.73 noro 2661: nd_nvar = nvar;
1.79 noro 2662: if ( max ) {
1.83 noro 2663: /* XXX */
2664: if ( do_weyl ) nd_bpe = 32;
2665: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 2666: else if ( max < 4 ) nd_bpe = 2;
2667: else if ( max < 8 ) nd_bpe = 3;
2668: else if ( max < 16 ) nd_bpe = 4;
2669: else if ( max < 32 ) nd_bpe = 5;
2670: else if ( max < 64 ) nd_bpe = 6;
2671: else if ( max < 256 ) nd_bpe = 8;
2672: else if ( max < 1024 ) nd_bpe = 10;
2673: else if ( max < 65536 ) nd_bpe = 16;
2674: else nd_bpe = 32;
2675: }
1.61 noro 2676: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2677: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2678: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2679: wpd = nd_exporigin+elen;
2680: if ( wpd != nd_wpd ) {
2681: nd_free_private_storage();
2682: nd_wpd = wpd;
2683: }
1.1 noro 2684: if ( nd_bpe < 32 ) {
2685: nd_mask0 = (1<<nd_bpe)-1;
2686: } else {
2687: nd_mask0 = 0xffffffff;
2688: }
2689: bzero(nd_mask,sizeof(nd_mask));
2690: nd_mask1 = 0;
2691: for ( i = 0; i < nd_epw; i++ ) {
2692: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2693: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2694: }
1.90 noro 2695: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 2696: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2697: nd_blockmask = nd_create_blockmask(nd_ord);
1.96 ! noro 2698: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 2699: }
2700:
1.20 noro 2701: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2702: {
1.37 noro 2703: int i,obpe,oadv,h;
1.71 noro 2704: static NM prev_nm_free_list;
2705: static ND_pairs prev_ndp_free_list;
1.13 noro 2706: RHist mr0,mr;
2707: RHist r;
1.37 noro 2708: RHist *old_red;
1.71 noro 2709: ND_pairs s0,s,t;
1.43 noro 2710: EPOS oepos;
1.15 noro 2711:
1.1 noro 2712: obpe = nd_bpe;
1.11 noro 2713: oadv = nmv_adv;
1.43 noro 2714: oepos = nd_epos;
1.63 noro 2715: if ( obpe < 2 ) nd_bpe = 2;
2716: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2717: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2718: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2719: else if ( obpe < 6 ) nd_bpe = 6;
2720: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2721: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2722: else if ( obpe < 16 ) nd_bpe = 16;
2723: else if ( obpe < 32 ) nd_bpe = 32;
2724: else error("nd_reconstruct : exponent too large");
1.5 noro 2725:
1.73 noro 2726: nd_setup_parameters(nd_nvar,0);
1.1 noro 2727: prev_nm_free_list = _nm_free_list;
2728: prev_ndp_free_list = _ndp_free_list;
2729: _nm_free_list = 0;
2730: _ndp_free_list = 0;
1.53 noro 2731: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2732: if ( trace )
2733: for ( i = nd_psn-1; i >= 0; i-- )
2734: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2735: s0 = 0;
2736: for ( t = d; t; t = NEXT(t) ) {
2737: NEXTND_pairs(s0,s);
2738: s->i1 = t->i1;
2739: s->i2 = t->i2;
1.14 noro 2740: SG(s) = SG(t);
1.61 noro 2741: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2742: }
1.37 noro 2743:
2744: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2745: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2746: old_red[i] = nd_red[i];
2747: nd_red[i] = 0;
2748: }
2749: for ( i = 0; i < REDTAB_LEN; i++ )
2750: for ( r = old_red[i]; r; r = NEXT(r) ) {
2751: NEWRHist(mr);
1.13 noro 2752: mr->index = r->index;
1.20 noro 2753: SG(mr) = SG(r);
1.61 noro 2754: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2755: h = ndl_hash_value(DL(mr));
2756: NEXT(mr) = nd_red[h];
2757: nd_red[h] = mr;
1.6 noro 2758: }
1.37 noro 2759: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2760: old_red = 0;
1.11 noro 2761: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2762: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2763: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2764: nd_psh[i] = r;
1.11 noro 2765: }
1.1 noro 2766: if ( s0 ) NEXT(s) = 0;
2767: prev_nm_free_list = 0;
2768: prev_ndp_free_list = 0;
1.71 noro 2769: #if 0
1.1 noro 2770: GC_gcollect();
1.71 noro 2771: #endif
1.1 noro 2772: return s0;
2773: }
2774:
1.61 noro 2775: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2776: {
1.57 noro 2777: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2778: struct order_pair *op;
1.1 noro 2779:
2780: n = nd_nvar;
1.61 noro 2781: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2782: omask0 = (1<<obpe)-1;
1.34 noro 2783: TD(r) = TD(d);
1.41 noro 2784: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2785: if ( nd_blockmask ) {
2786: l = nd_blockmask->n;
2787: op = nd_blockmask->order_pair;
2788: for ( i = 1; i < nd_exporigin; i++ )
2789: r[i] = d[i];
2790: for ( j = 0, s = 0; j < l; j++ ) {
2791: ord_l = op[j].length;
1.57 noro 2792: for ( i = 0; i < ord_l; i++, s++ ) {
2793: ei = GET_EXP_OLD(d,s);
2794: PUT_EXP(r,s,ei);
2795: }
1.1 noro 2796: }
1.43 noro 2797: } else {
1.56 noro 2798: for ( i = 0; i < n; i++ ) {
2799: ei = GET_EXP_OLD(d,i);
2800: PUT_EXP(r,i,ei);
2801: }
1.1 noro 2802: }
2803: }
1.3 noro 2804:
1.6 noro 2805: ND nd_copy(ND p)
2806: {
2807: NM m,mr,mr0;
1.41 noro 2808: int c,n;
1.6 noro 2809: ND r;
2810:
2811: if ( !p )
2812: return 0;
2813: else {
2814: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2815: NEXTNM(mr0,mr);
1.14 noro 2816: CM(mr) = CM(m);
2817: ndl_copy(DL(m),DL(mr));
1.6 noro 2818: }
2819: NEXT(mr) = 0;
1.31 noro 2820: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2821: SG(r) = SG(p);
1.6 noro 2822: return r;
2823: }
2824: }
2825:
1.53 noro 2826: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2827: {
2828: NM m;
2829: NDV p1,p2;
2830: ND t1,t2;
1.61 noro 2831: UINT *lcm;
1.31 noro 2832: int td;
1.11 noro 2833:
1.77 noro 2834: if ( !mod && nd_demand ) {
2835: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 2836: } else {
1.77 noro 2837: if ( trace ) {
2838: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
2839: } else {
2840: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
2841: }
1.20 noro 2842: }
1.34 noro 2843: lcm = LCM(p);
1.11 noro 2844: NEWNM(m);
1.20 noro 2845: CQ(m) = HCQ(p2);
1.34 noro 2846: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 2847: if ( ndl_check_bound2(p->i1,DL(m)) )
2848: return 0;
1.55 noro 2849: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 2850: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
2851: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 2852: else chsgnq(HCQ(p1),&CQ(m));
2853: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 2854: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 2855: nd_free(t1);
2856: return 0;
2857: }
1.55 noro 2858: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 2859: *rp = nd_add(mod,t1,t2);
1.11 noro 2860: FREENM(m);
2861: return 1;
2862: }
2863:
1.19 noro 2864: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 2865: {
2866: NMV m;
2867: int c,c1,len,i;
2868:
1.34 noro 2869: if ( !p ) return;
1.14 noro 2870: len = LEN(p);
1.71 noro 2871: if ( mod == -1 )
2872: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
2873: CM(m) = _mulsf(CM(m),mul);
2874: else
2875: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2876: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2877: }
1.11 noro 2878: }
2879:
1.16 noro 2880: void ndv_mul_c_q(NDV p,Q mul)
2881: {
2882: NMV m;
2883: Q c;
2884: int len,i;
2885:
1.34 noro 2886: if ( !p ) return;
1.16 noro 2887: len = LEN(p);
2888: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2889: mulq(CQ(m),mul,&c); CQ(m) = c;
2890: }
2891: }
2892:
1.55 noro 2893: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
2894: int n2,i,j,l,n,tlen;
1.61 noro 2895: UINT *d0;
1.55 noro 2896: NM *tab,*psum;
2897: ND s,r;
2898: NM t;
2899: NMV m1;
2900:
2901: if ( !p ) return 0;
2902: n = NV(p); n2 = n>>1;
2903: d0 = DL(m0);
2904: l = LEN(p);
2905: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
2906: tab = (NM *)ALLOCA(tlen*sizeof(NM));
2907: psum = (NM *)ALLOCA(tlen*sizeof(NM));
2908: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 2909: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
2910: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 2911: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 2912: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 2913: for ( j = 0; j < tlen; j++ ) {
2914: if ( tab[j] ) {
2915: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
2916: }
2917: }
2918: }
2919: for ( i = tlen-1, r = 0; i >= 0; i-- )
2920: if ( psum[i] ) {
2921: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
2922: MKND(n,psum[i],j,s);
2923: r = nd_add(mod,r,s);
2924: }
1.56 noro 2925: if ( r ) SG(r) = SG(p)+TD(d0);
2926: return r;
1.55 noro 2927: }
2928:
1.56 noro 2929: /* product of monomials */
2930: /* XXX block order is not handled correctly */
2931:
1.55 noro 2932: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
2933: {
1.56 noro 2934: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 2935: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 2936: Q *ctab_q;
2937: Q q,q1;
1.61 noro 2938: UINT c0,c1,c;
1.55 noro 2939: NM *p;
2940: NM m,t;
2941:
2942: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
2943: if ( !m0 || !m1 ) return;
2944: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
2945: NEWNM(m); d = DL(m);
1.56 noro 2946: if ( mod ) {
2947: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
2948: } else
2949: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 2950: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
2951: homo = n&1 ? 1 : 0;
2952: if ( homo ) {
2953: /* offset of h-degree */
2954: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
2955: PUT_EXP(DL(m),n-1,h);
2956: TD(DL(m)) = h;
1.61 noro 2957: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 2958: }
2959: tab[0] = m;
2960: NEWNM(m); d = DL(m);
1.57 noro 2961: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 2962: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
2963: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
2964: /* xi^a*(Di^k*xi^l)*Di^b */
2965: a += l; b += k;
1.56 noro 2966: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 2967: if ( !k || !l ) {
2968: for ( j = 0; j < curlen; j++ )
1.56 noro 2969: if ( t = tab[j] ) {
2970: dt = DL(t);
2971: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 2972: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 2973: }
2974: curlen *= k+1;
2975: continue;
2976: }
2977: min = MIN(k,l);
1.56 noro 2978: if ( mod ) {
1.61 noro 2979: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 2980: mkwcm(k,l,mod,ctab);
2981: } else {
2982: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
2983: mkwc(k,l,ctab_q);
2984: }
1.57 noro 2985: for ( j = min; j >= 0; j-- ) {
1.56 noro 2986: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 2987: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 2988: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 2989: if ( homo ) {
2990: TD(d) = s;
1.56 noro 2991: PUT_EXP(d,n-1,s-h);
1.55 noro 2992: } else TD(d) = h;
1.61 noro 2993: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 2994: if ( mod ) c = ctab[j];
2995: else q = ctab_q[j];
1.57 noro 2996: p = tab+curlen*j;
2997: if ( j == 0 ) {
2998: for ( u = 0; u < curlen; u++, p++ ) {
2999: if ( tab[u] ) {
3000: ndl_addto(DL(tab[u]),d);
3001: if ( mod ) {
3002: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3003: } else {
3004: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3005: }
3006: }
1.56 noro 3007: }
1.57 noro 3008: } else {
3009: for ( u = 0; u < curlen; u++, p++ ) {
3010: if ( tab[u] ) {
3011: NEWNM(t);
3012: ndl_add(DL(tab[u]),d,DL(t));
3013: if ( mod ) {
3014: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3015: } else
3016: mulq(CQ(tab[u]),q,&CQ(t));
3017: *p = t;
3018: }
1.55 noro 3019: }
3020: }
3021: }
3022: curlen *= k+1;
3023: }
3024: FREENM(m);
3025: }
3026:
1.63 noro 3027: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3028: {
3029: NM mr,mr0;
3030: NMV m;
3031: UINT *d,*dt,*dm;
3032: int c,n,td,i,c1,c2,len;
3033: Q q;
3034: ND r;
3035:
3036: if ( !p ) return 0;
3037: else {
3038: n = NV(p); m = BDY(p);
3039: d = DL(m0);
3040: len = LEN(p);
3041: mr0 = 0;
3042: td = TD(d);
3043: c = CM(m0);
3044: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3045: NEXTNM(mr0,mr);
3046: CM(mr) = 1;
3047: ndl_add(DL(m),d,DL(mr));
3048: }
3049: NEXT(mr) = 0;
3050: MKND(NV(p),mr0,len,r);
3051: SG(r) = SG(p) + TD(d);
3052: return r;
3053: }
3054: }
3055:
1.55 noro 3056: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3057: {
3058: NM mr,mr0;
3059: NMV m;
1.61 noro 3060: UINT *d,*dt,*dm;
1.9 noro 3061: int c,n,td,i,c1,c2,len;
1.16 noro 3062: Q q;
1.9 noro 3063: ND r;
3064:
1.34 noro 3065: if ( !p ) return 0;
1.55 noro 3066: else if ( do_weyl )
1.71 noro 3067: if ( mod == -1 )
3068: error("ndv_mul_nm : not implemented (weyl)");
3069: else
3070: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3071: else {
3072: n = NV(p); m = BDY(p);
1.34 noro 3073: d = DL(m0);
1.14 noro 3074: len = LEN(p);
1.9 noro 3075: mr0 = 0;
1.34 noro 3076: td = TD(d);
1.71 noro 3077: if ( mod == -1 ) {
3078: c = CM(m0);
3079: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3080: NEXTNM(mr0,mr);
3081: CM(mr) = _mulsf(CM(m),c);
3082: ndl_add(DL(m),d,DL(mr));
3083: }
3084: } else if ( mod ) {
1.16 noro 3085: c = CM(m0);
3086: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3087: NEXTNM(mr0,mr);
3088: c1 = CM(m);
1.19 noro 3089: DMAR(c1,c,0,mod,c2);
1.16 noro 3090: CM(mr) = c2;
3091: ndl_add(DL(m),d,DL(mr));
3092: }
3093: } else {
3094: q = CQ(m0);
3095: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3096: NEXTNM(mr0,mr);
3097: mulq(CQ(m),q,&CQ(mr));
3098: ndl_add(DL(m),d,DL(mr));
3099: }
1.4 noro 3100: }
1.9 noro 3101: NEXT(mr) = 0;
1.31 noro 3102: MKND(NV(p),mr0,len,r);
1.34 noro 3103: SG(r) = SG(p) + TD(d);
1.9 noro 3104: return r;
1.4 noro 3105: }
3106: }
3107:
1.43 noro 3108: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3109: {
1.13 noro 3110: NMV m,mr,mr0,t;
3111: int len,i,k;
1.11 noro 3112:
1.61 noro 3113: if ( !p ) return;
3114: m = BDY(p); len = LEN(p);
3115: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3116: m = (NMV)((char *)mr0+(len-1)*oadv);
3117: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3118: t = (NMV)ALLOCA(nmv_adv);
3119: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3120: CQ(t) = CQ(m);
3121: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3122: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3123: CQ(mr) = CQ(t);
3124: ndl_copy(DL(t),DL(mr));
3125: }
3126: BDY(p) = mr0;
3127: }
3128:
3129: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3130: {
3131: NMV m,mr,mr0;
3132: int len,i;
3133: NDV r;
1.11 noro 3134:
1.61 noro 3135: if ( !p ) return 0;
3136: m = BDY(p); len = LEN(p);
3137: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3138: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3139: ndl_zero(DL(mr));
3140: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3141: CQ(mr) = CQ(m);
1.11 noro 3142: }
1.61 noro 3143: MKNDV(NV(p),mr0,len,r);
3144: SG(r) = SG(p);
3145: return r;
1.11 noro 3146: }
3147:
1.61 noro 3148: /* duplicate p */
3149:
3150: NDV ndv_dup(int mod,NDV p)
1.3 noro 3151: {
3152: NDV d;
1.61 noro 3153: NMV t,m,m0;
1.3 noro 3154: int i,len;
3155:
1.34 noro 3156: if ( !p ) return 0;
1.31 noro 3157: len = LEN(p);
1.34 noro 3158: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3159: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3160: ndl_copy(DL(t),DL(m));
1.16 noro 3161: CQ(m) = CQ(t);
1.3 noro 3162: }
3163: MKNDV(NV(p),m0,len,d);
1.23 noro 3164: SG(d) = SG(p);
3165: return d;
3166: }
3167:
1.63 noro 3168: ND nd_dup(ND p)
3169: {
3170: ND d;
3171: NM t,m,m0;
3172:
3173: if ( !p ) return 0;
3174: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3175: NEXTNM(m0,m);
3176: ndl_copy(DL(t),DL(m));
3177: CQ(m) = CQ(t);
3178: }
3179: if ( m0 ) NEXT(m) = 0;
3180: MKND(NV(p),m0,LEN(p),d);
3181: SG(d) = SG(p);
3182: return d;
3183: }
3184:
1.61 noro 3185: /* XXX if p->len == 0 then it represents 0 */
3186:
3187: void ndv_mod(int mod,NDV p)
3188: {
3189: NMV t,d;
3190: int r;
3191: int i,len,dlen;
1.71 noro 3192: Obj gfs;
1.61 noro 3193:
3194: if ( !p ) return;
3195: len = LEN(p);
3196: dlen = 0;
1.71 noro 3197: if ( mod == -1 )
3198: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3199: simp_ff((Obj)CP(t),&gfs);
3200: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3201: CM(d) = r;
3202: ndl_copy(DL(t),DL(d));
3203: NMV_ADV(d);
3204: dlen++;
3205: }
1.71 noro 3206: else
3207: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3208: r = rem(NM(CQ(t)),mod);
3209: if ( r ) {
3210: if ( SGN(CQ(t)) < 0 )
3211: r = mod-r;
3212: CM(d) = r;
3213: ndl_copy(DL(t),DL(d));
3214: NMV_ADV(d);
3215: dlen++;
3216: }
3217: }
1.61 noro 3218: LEN(p) = dlen;
3219: }
3220:
3221: NDV ptondv(VL vl,VL dvl,P p)
3222: {
3223: ND nd;
3224:
3225: nd = ptond(vl,dvl,p);
3226: return ndtondv(0,nd);
3227: }
3228:
3229: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3230: {
1.61 noro 3231: int n,i,j,k,e;
3232: VL tvl;
3233: V v;
3234: DCP dc;
3235: DCP *w;
3236: ND r,s,t,u;
3237: P x;
3238: int c;
3239: UINT *d;
1.23 noro 3240: NM m,m0;
1.61 noro 3241:
3242: if ( !p )
3243: return 0;
3244: else if ( NUM(p) ) {
3245: NEWNM(m);
3246: ndl_zero(DL(m));
3247: CQ(m) = (Q)p;
3248: NEXT(m) = 0;
3249: MKND(nd_nvar,m,1,r);
3250: SG(r) = 0;
3251: return r;
3252: } else {
3253: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3254: w = (DCP *)ALLOCA(k*sizeof(DCP));
3255: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3256: for ( i = 0, tvl = dvl, v = VR(p);
3257: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3258: if ( !tvl ) {
3259: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3260: t = ptond(vl,dvl,COEF(w[j]));
3261: pwrp(vl,x,DEG(w[j]),&p);
3262: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3263: }
3264: return s;
3265: } else {
3266: NEWNM(m0); d = DL(m0);
3267: for ( j = k-1, s = 0; j >= 0; j-- ) {
3268: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3269: TD(d) = MUL_WEIGHT(e,i);
3270: if ( nd_blockmask) ndl_weight_mask(d);
3271: t = ptond(vl,dvl,COEF(w[j]));
3272: for ( m = BDY(t); m; m = NEXT(m) )
3273: ndl_addto(DL(m),d);
3274: SG(t) += TD(d);
3275: s = nd_add(0,s,t);
3276: }
3277: FREENM(m0);
3278: return s;
3279: }
3280: }
3281: }
3282:
3283: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3284: {
3285: VL tvl;
3286: int len,n,j,i,e;
3287: NMV m;
3288: Q q;
3289: P c;
3290: UINT *d;
3291: P s,r,u,t,w;
1.71 noro 3292: GFS gfs;
1.23 noro 3293:
1.34 noro 3294: if ( !p ) return 0;
1.61 noro 3295: else {
3296: len = LEN(p);
3297: n = NV(p);
3298: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3299: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3300: if ( mod == -1 ) {
3301: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3302: } else if ( mod ) {
1.61 noro 3303: STOQ(CM(m),q); c = (P)q;
3304: } else
3305: c = CP(m);
3306: d = DL(m);
3307: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3308: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3309: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3310: }
3311: addp(vl,s,t,&u); s = u;
3312: }
3313: return s;
1.23 noro 3314: }
1.3 noro 3315: }
3316:
1.61 noro 3317: NDV ndtondv(int mod,ND p)
1.11 noro 3318: {
3319: NDV d;
1.61 noro 3320: NMV m,m0;
3321: NM t;
3322: int i,len;
1.11 noro 3323:
1.34 noro 3324: if ( !p ) return 0;
1.61 noro 3325: len = LEN(p);
1.74 noro 3326: if ( mod )
3327: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3328: else
3329: m0 = m = MALLOC(len*nmv_adv);
3330: ndv_alloc += nmv_adv*len;
1.61 noro 3331: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3332: ndl_copy(DL(t),DL(m));
3333: CQ(m) = CQ(t);
1.11 noro 3334: }
1.61 noro 3335: MKNDV(NV(p),m0,len,d);
1.14 noro 3336: SG(d) = SG(p);
1.11 noro 3337: return d;
3338: }
3339:
1.61 noro 3340: ND ndvtond(int mod,NDV p)
1.11 noro 3341: {
1.61 noro 3342: ND d;
3343: NM m,m0;
1.11 noro 3344: NMV t;
1.61 noro 3345: int i,len;
1.11 noro 3346:
1.34 noro 3347: if ( !p ) return 0;
1.11 noro 3348: m0 = 0;
1.61 noro 3349: len = p->len;
3350: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3351: NEXTNM(m0,m);
3352: ndl_copy(DL(t),DL(m));
3353: CQ(m) = CQ(t);
1.11 noro 3354: }
3355: NEXT(m) = 0;
1.61 noro 3356: MKND(NV(p),m0,len,d);
1.14 noro 3357: SG(d) = SG(p);
1.11 noro 3358: return d;
3359: }
3360:
1.3 noro 3361: void ndv_print(NDV p)
3362: {
3363: NMV m;
3364: int i,len;
3365:
1.34 noro 3366: if ( !p ) printf("0\n");
1.3 noro 3367: else {
1.14 noro 3368: len = LEN(p);
1.3 noro 3369: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3370: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3371: else printf("+%d*",CM(m));
1.16 noro 3372: ndl_print(DL(m));
3373: }
3374: printf("\n");
3375: }
3376: }
3377:
3378: void ndv_print_q(NDV p)
3379: {
3380: NMV m;
3381: int i,len;
3382:
1.34 noro 3383: if ( !p ) printf("0\n");
1.16 noro 3384: else {
3385: len = LEN(p);
3386: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3387: printf("+");
1.74 noro 3388: printexpr(CO,(Obj)CQ(m));
1.16 noro 3389: printf("*");
1.14 noro 3390: ndl_print(DL(m));
1.3 noro 3391: }
3392: printf("\n");
3393: }
1.25 noro 3394: }
3395:
1.61 noro 3396: NODE ndv_reducebase(NODE x)
1.27 noro 3397: {
3398: int len,i,j;
3399: NDV *w;
3400: NODE t,t0;
3401:
3402: len = length(x);
3403: w = (NDV *)ALLOCA(len*sizeof(NDV));
3404: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3405: for ( i = 0; i < len; i++ ) {
3406: for ( j = 0; j < i; j++ ) {
3407: if ( w[i] && w[j] )
3408: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3409: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3410: }
3411: }
3412: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3413: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3414: }
3415: NEXT(t) = 0; x = t0;
3416: return x;
1.11 noro 3417: }
1.32 noro 3418:
1.43 noro 3419: /* XXX incomplete */
3420:
1.32 noro 3421: void nd_init_ord(struct order_spec *ord)
3422: {
1.43 noro 3423: switch ( ord->id ) {
1.32 noro 3424: case 0:
1.43 noro 3425: switch ( ord->ord.simple ) {
3426: case 0:
3427: nd_dcomp = 1;
3428: nd_isrlex = 1;
3429: break;
3430: case 1:
3431: nd_dcomp = 1;
3432: nd_isrlex = 0;
3433: break;
3434: case 2:
3435: nd_dcomp = 0;
3436: nd_isrlex = 0;
1.45 noro 3437: ndl_compare_function = ndl_lex_compare;
1.58 noro 3438: break;
3439: case 11:
3440: /* XXX */
3441: nd_dcomp = 0;
3442: nd_isrlex = 1;
3443: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3444: break;
3445: default:
3446: error("nd_gr : unsupported order");
3447: }
1.32 noro 3448: break;
3449: case 1:
1.96 ! noro 3450: /* block order */
1.43 noro 3451: /* XXX */
3452: nd_dcomp = -1;
1.32 noro 3453: nd_isrlex = 0;
1.45 noro 3454: ndl_compare_function = ndl_block_compare;
1.34 noro 3455: break;
1.43 noro 3456: case 2:
1.96 ! noro 3457: /* matrix order */
! 3458: /* XXX */
! 3459: nd_dcomp = -1;
! 3460: nd_isrlex = 0;
! 3461: nd_matrix_len = ord->ord.matrix.row;
! 3462: nd_matrix = ord->ord.matrix.matrix;
! 3463: ndl_compare_function = ndl_matrix_compare;
! 3464: break;
! 3465: case 3:
! 3466: error("nd_init_ord : composite order is not supported yet.");
1.32 noro 3467: break;
3468: }
1.41 noro 3469: nd_ord = ord;
1.32 noro 3470: }
3471:
1.43 noro 3472: BlockMask nd_create_blockmask(struct order_spec *ord)
3473: {
3474: int n,i,j,s,l;
1.61 noro 3475: UINT *t;
1.43 noro 3476: BlockMask bm;
3477:
1.96 ! noro 3478: /* we only create mask table for block order */
! 3479: if ( ord->id != 1 )
1.43 noro 3480: return 0;
3481: n = ord->ord.block.length;
3482: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3483: bm->n = n;
3484: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3485: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3486: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3487: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3488: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3489: l = bm->order_pair[i].length;
3490: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3491: }
3492: return bm;
1.57 noro 3493: }
3494:
3495: EPOS nd_create_epos(struct order_spec *ord)
3496: {
3497: int i,j,l,s,ord_l,ord_o;
3498: EPOS epos;
3499: struct order_pair *op;
3500:
3501: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3502: switch ( ord->id ) {
3503: case 0:
3504: if ( nd_isrlex ) {
3505: for ( i = 0; i < nd_nvar; i++ ) {
3506: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3507: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3508: }
3509: } else {
3510: for ( i = 0; i < nd_nvar; i++ ) {
3511: epos[i].i = nd_exporigin + i/nd_epw;
3512: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3513: }
3514: }
3515: break;
3516: case 1:
3517: /* block order */
3518: l = ord->ord.block.length;
3519: op = ord->ord.block.order_pair;
3520: for ( j = 0, s = 0; j < l; j++ ) {
3521: ord_o = op[j].order;
3522: ord_l = op[j].length;
3523: if ( !ord_o )
3524: for ( i = 0; i < ord_l; i++ ) {
3525: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3526: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3527: }
3528: else
3529: for ( i = 0; i < ord_l; i++ ) {
3530: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3531: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3532: }
3533: s += ord_l;
3534: }
3535: break;
3536: case 2:
1.96 ! noro 3537: /* matrix order */
! 3538: case 3:
! 3539: /* composite order */
! 3540: for ( i = 0; i < nd_nvar; i++ ) {
! 3541: epos[i].i = nd_exporigin + i/nd_epw;
! 3542: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
! 3543: }
! 3544: break;
1.57 noro 3545: }
3546: return epos;
1.43 noro 3547: }
1.59 noro 3548:
3549: /* external interface */
3550:
3551: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3552: {
1.61 noro 3553: NODE t,in0,in;
1.59 noro 3554: ND nd,nf;
1.61 noro 3555: NDV ndv;
3556: VL vv,tv;
3557: int stat,nvar,max,e;
1.69 noro 3558: union oNDC dn;
1.59 noro 3559:
3560: pltovl(v,&vv);
1.74 noro 3561: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3562:
3563: /* get the degree bound */
3564: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3565: for ( tv = vv; tv; tv = NEXT(tv) ) {
3566: e = getdeg(tv->v,(P)BDY(t));
3567: max = MAX(e,max);
3568: }
3569: for ( tv = vv; tv; tv = NEXT(tv) ) {
3570: e = getdeg(tv->v,f);
3571: max = MAX(e,max);
3572: }
3573:
1.59 noro 3574: nd_init_ord(ord);
1.61 noro 3575: nd_setup_parameters(nvar,max);
3576:
3577: /* conversion to ndv */
3578: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3579: NEXTNODE(in0,in);
3580: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3581: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3582: }
3583: NEXTNODE(in0,in);
3584: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3585: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3586: NEXT(in) = 0;
3587:
3588: ndv_setup(m,0,in0);
1.59 noro 3589: nd_psn--;
3590: nd_scale=2;
3591: while ( 1 ) {
3592: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3593: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3594: if ( !stat ) {
3595: nd_psn++;
3596: nd_reconstruct(m,0,0);
3597: nd_psn--;
3598: } else
3599: break;
3600: }
1.61 noro 3601: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3602: }
3603:
3604: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3605: {
3606: NM m;
3607: UINT *t,*s;
3608: int i;
3609:
3610: for ( i = 0; i < n; i++ ) r[i] = 0;
3611: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3612: t = DL(m);
3613: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3614: r[i] = CM(m);
3615: }
3616: for ( i = 0; !r[i]; i++ );
3617: return i;
3618: }
3619:
1.74 noro 3620: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3621: {
3622: NMV m;
3623: UINT *t,*s;
3624: int i,j,len;
3625:
3626: for ( i = 0; i < n; i++ ) r[i] = 0;
3627: m = BDY(d);
3628: len = LEN(d);
3629: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3630: t = DL(m);
3631: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3632: r[i] = CM(m);
3633: }
3634: for ( i = 0; !r[i]; i++ );
3635: return i;
3636: }
3637:
1.63 noro 3638: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3639: {
3640: NM m;
3641: NMV mr;
3642: UINT *d,*t,*s;
3643: NDV p;
3644: int i,j,len;
3645:
3646: m = pair->mul;
3647: d = DL(m);
3648: p = nd_ps[pair->index];
3649: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3650: for ( i = 0; i < n; i++ ) r[i] = 0;
3651: len = LEN(p);
3652: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3653: ndl_add(d,DL(mr),t);
3654: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3655: r[i] = CM(mr);
3656: }
3657: for ( i = 0; !r[i]; i++ );
3658: return i;
3659: }
3660:
1.67 noro 3661: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3662: {
3663: NM m;
3664: NMV mr;
3665: UINT *d,*t,*s;
3666: NDV p;
1.67 noro 3667: unsigned char *ivc;
3668: unsigned short *ivs;
1.81 noro 3669: UINT *v,*ivi,*s0v;
1.67 noro 3670: int i,j,len,prev,diff,cdiff;
3671: IndArray r;
1.64 noro 3672:
3673: m = pair->mul;
3674: d = DL(m);
3675: p = nd_ps[pair->index];
3676: len = LEN(p);
3677: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3678: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3679: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3680: ndl_add(d,DL(mr),t);
3681: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3682: v[j] = i;
3683: }
1.81 noro 3684: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 3685: r->head = v[0];
3686: diff = 0;
3687: for ( i = 1; i < len; i++ ) {
3688: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3689: }
3690: if ( diff < 256 ) {
3691: r->width = 1;
3692: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3693: r->index.c = ivc;
3694: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3695: } else if ( diff < 65536 ) {
3696: r->width = 2;
3697: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3698: r->index.s = ivs;
3699: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3700: } else {
3701: r->width = 4;
3702: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3703: r->index.i = ivi;
3704: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3705: }
1.67 noro 3706: return r;
1.64 noro 3707: }
3708:
1.65 noro 3709:
1.76 noro 3710: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3711: {
1.67 noro 3712: int i,j,k,len,pos,prev;
1.66 noro 3713: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3714: IndArray ivect;
3715: unsigned char *ivc;
3716: unsigned short *ivs;
3717: unsigned int *ivi;
1.65 noro 3718: NDV redv;
1.67 noro 3719: NMV mr;
1.65 noro 3720: NODE rp;
1.76 noro 3721: int maxrs;
1.65 noro 3722:
1.76 noro 3723: maxrs = 0;
1.74 noro 3724: for ( i = 0; i < nred; i++ ) {
1.65 noro 3725: ivect = imat[i];
1.67 noro 3726: k = ivect->head; svect[k] %= m;
1.65 noro 3727: if ( c = svect[k] ) {
1.76 noro 3728: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3729: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3730: len = LEN(redv); mr = BDY(redv);
3731: svect[k] = 0; prev = k;
3732: switch ( ivect->width ) {
3733: case 1:
3734: ivc = ivect->index.c;
3735: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3736: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3737: prev = pos;
3738: DMA(c1,c,c2,up,lo);
3739: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3740: } else svect[pos] = lo;
3741: }
3742: break;
3743: case 2:
3744: ivs = ivect->index.s;
3745: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3746: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3747: prev = pos;
3748: DMA(c1,c,c2,up,lo);
3749: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3750: } else svect[pos] = lo;
3751: }
3752: break;
3753: case 4:
3754: ivi = ivect->index.i;
3755: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3756: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3757: prev = pos;
3758: DMA(c1,c,c2,up,lo);
3759: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3760: } else svect[pos] = lo;
3761: }
3762: break;
1.65 noro 3763: }
3764: }
3765: }
1.66 noro 3766: for ( i = 0; i < col; i++ )
3767: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3768: return maxrs;
1.65 noro 3769: }
3770:
1.76 noro 3771: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3772: {
3773: int i,j,k,len,pos,prev;
3774: UINT c,c1,c2,c3,up,lo,dmy;
3775: IndArray ivect;
3776: unsigned char *ivc;
3777: unsigned short *ivs;
3778: unsigned int *ivi;
3779: NDV redv;
3780: NMV mr;
3781: NODE rp;
1.76 noro 3782: int maxrs;
1.72 noro 3783:
1.76 noro 3784: maxrs = 0;
1.74 noro 3785: for ( i = 0; i < nred; i++ ) {
1.72 noro 3786: ivect = imat[i];
3787: k = ivect->head; svect[k] %= m;
3788: if ( c = svect[k] ) {
1.76 noro 3789: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3790: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3791: len = LEN(redv); mr = BDY(redv);
3792: svect[k] = 0; prev = k;
3793: switch ( ivect->width ) {
3794: case 1:
3795: ivc = ivect->index.c;
3796: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3797: pos = prev+ivc[j]; prev = pos;
3798: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3799: }
3800: break;
3801: case 2:
3802: ivs = ivect->index.s;
3803: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3804: pos = prev+ivs[j]; prev = pos;
3805: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3806: }
3807: break;
3808: case 4:
3809: ivi = ivect->index.i;
3810: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3811: pos = prev+ivi[j]; prev = pos;
3812: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3813: }
3814: break;
3815: }
3816: }
3817: }
1.76 noro 3818: return maxrs;
1.72 noro 3819: }
3820:
1.65 noro 3821: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3822: {
3823: int j,k,len;
3824: UINT *p;
3825: UINT c;
3826: NDV r;
3827: NMV mr0,mr;
3828:
3829: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3830: if ( !len ) return 0;
3831: else {
1.74 noro 3832: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
3833: ndv_alloc += nmv_adv*len;
1.65 noro 3834: mr = mr0;
3835: p = s0vect;
3836: for ( j = k = 0; j < col; j++, p += nd_wpd )
3837: if ( !rhead[j] ) {
3838: if ( c = vect[k++] ) {
3839: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3840: }
3841: }
3842: MKNDV(nd_nvar,mr0,len,r);
3843: return r;
3844: }
3845: }
3846:
1.74 noro 3847: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 3848: {
3849: ND_pairs t;
3850: NODE sp0,sp;
3851: int stat;
3852: ND spol;
3853:
3854: for ( t = l; t; t = NEXT(t) ) {
3855: stat = nd_sp(m,0,t,&spol);
3856: if ( !stat ) return 0;
3857: if ( spol ) {
3858: add_pbucket_symbolic(bucket,spol);
3859: }
3860: }
1.68 noro 3861: return 1;
1.65 noro 3862: }
3863:
3864: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
3865: {
3866: NODE rp0,rp;
3867: NM mul,head,s0,s;
1.76 noro 3868: int index,col,i,sugar;
1.65 noro 3869: RHist h;
3870: UINT *s0v,*p;
3871: NM_ind_pair pair;
3872: ND red;
3873:
3874: s0 = 0; rp0 = 0; col = 0;
3875: while ( 1 ) {
3876: head = remove_head_pbucket_symbolic(bucket);
3877: if ( !head ) break;
3878: if ( !s0 ) s0 = head;
3879: else NEXT(s) = head;
3880: s = head;
3881: index = ndl_find_reducer(DL(head));
3882: if ( index >= 0 ) {
3883: h = nd_psh[index];
3884: NEWNM(mul);
3885: ndl_sub(DL(head),DL(h),DL(mul));
3886: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 3887: sugar = TD(DL(mul))+SG(nd_ps[index]);
3888: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 3889: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
3890: add_pbucket_symbolic(bucket,nd_remove_head(red));
3891: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
3892: }
3893: col++;
3894: }
1.72 noro 3895: if ( rp0 ) NEXT(rp) = 0;
3896: NEXT(s) = 0;
1.65 noro 3897: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
3898: for ( i = 0, p = s0v, s = s0; i < col;
3899: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
3900: *s0vect = s0v;
3901: *r = rp0;
3902: return col;
3903: }
3904:
1.69 noro 3905: NODE nd_f4(int m)
3906: {
3907: int i,nh,stat,index;
3908: NODE r,g;
3909: ND_pairs d,l,t;
3910: ND spol,red;
3911: NDV nf,redv;
3912: NM s0,s;
1.84 noro 3913: NODE rp0,srp0,nflist;
1.69 noro 3914: int nsp,nred,col,rank,len,k,j,a;
3915: UINT c;
1.74 noro 3916: UINT **spmat;
1.69 noro 3917: UINT *s0vect,*svect,*p,*v;
3918: int *colstat;
3919: IndArray *imat;
3920: int *rhead;
3921: int spcol,sprow;
3922: int sugar;
3923: PGeoBucket bucket;
3924: struct oEGT eg0,eg1,eg_f4;
3925:
3926: if ( !m )
3927: error("nd_f4 : not implemented");
1.74 noro 3928: ndv_alloc = 0;
1.69 noro 3929: g = 0; d = 0;
3930: for ( i = 0; i < nd_psn; i++ ) {
3931: d = update_pairs(d,g,i);
3932: g = update_base(g,i);
3933: }
3934: while ( d ) {
3935: get_eg(&eg0);
3936: l = nd_minsugarp(d,&d);
3937: sugar = SG(l);
3938: bucket = create_pbucket();
1.74 noro 3939: stat = nd_sp_f4(m,l,bucket);
1.69 noro 3940: if ( !stat ) {
3941: for ( t = l; NEXT(t); t = NEXT(t) );
3942: NEXT(t) = d; d = l;
3943: d = nd_reconstruct(m,0,d);
3944: continue;
3945: }
1.86 noro 3946: if ( bucket->m < 0 ) continue;
1.69 noro 3947: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
3948: if ( !col ) {
3949: for ( t = l; NEXT(t); t = NEXT(t) );
3950: NEXT(t) = d; d = l;
3951: d = nd_reconstruct(m,0,d);
3952: continue;
3953: }
3954: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 3955: if ( DP_Print )
3956: fprintf(asir_out,"sugar=%d,symb=%fsec,",
3957: sugar,eg_f4.exectime+eg_f4.gctime);
3958: if ( 1 )
3959: nflist = nd_f4_red(m,l,s0vect,col,rp0);
3960: else
3961: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 3962: /* adding new bases */
1.74 noro 3963: for ( r = nflist; r; r = NEXT(r) ) {
3964: nf = (NDV)BDY(r);
1.69 noro 3965: ndv_removecont(m,nf);
1.77 noro 3966: nh = ndv_newps(m,nf,0);
1.69 noro 3967: d = update_pairs(d,g,nh);
3968: g = update_base(g,nh);
3969: }
3970: }
3971: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 3972: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 3973: return g;
3974: }
1.74 noro 3975:
3976: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 3977: {
1.67 noro 3978: IndArray *imat;
1.74 noro 3979: int nsp,nred,spcol,sprow,a;
1.65 noro 3980: int *rhead;
1.74 noro 3981: int i,j,k,l,rank;
3982: NODE rp,r0,r;
3983: ND_pairs sp;
3984: ND spol;
3985: int **spmat;
3986: UINT *svect,*v;
3987: int *colstat;
1.80 noro 3988: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
1.74 noro 3989: NM_ind_pair *rvect;
1.76 noro 3990: int maxrs;
3991: int *spsugar;
1.63 noro 3992:
1.74 noro 3993: get_eg(&eg0);
3994: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
3995: nred = length(rp0); spcol = col-nred;
3996: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
3997: rhead = (int *)ALLOCA(col*sizeof(int));
3998: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 3999:
1.74 noro 4000: /* construction of index arrays */
4001: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4002: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4003: rvect[i] = (NM_ind_pair)BDY(rp);
4004: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4005: rhead[imat[i]->head] = 1;
4006: }
4007:
4008: /* elimination (1st step) */
4009: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4010: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4011: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4012: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4013: nd_sp(m,0,sp,&spol);
1.75 noro 4014: if ( !spol ) continue;
1.74 noro 4015: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4016: if ( m == -1 )
4017: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4018: else
4019: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4020: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4021: if ( i < col ) {
4022: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4023: for ( j = k = 0; j < col; j++ )
4024: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4025: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4026: sprow++;
4027: }
1.76 noro 4028: nd_free(spol);
1.74 noro 4029: }
1.80 noro 4030: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4031: if ( DP_Print ) {
4032: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4033: fflush(asir_out);
4034: }
1.74 noro 4035: /* free index arrays */
4036: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4037:
4038: /* elimination (2nd step) */
4039: colstat = (int *)ALLOCA(spcol*sizeof(int));
4040: if ( m == -1 )
1.76 noro 4041: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4042: else
1.76 noro 4043: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4044: r0 = 0;
4045: for ( i = 0; i < rank; i++ ) {
4046: NEXTNODE(r0,r); BDY(r) =
4047: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4048: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4049: GC_free(spmat[i]);
4050: }
4051: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4052: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4053: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4054: if ( DP_Print ) {
1.80 noro 4055: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4056: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4057: nsp,nred,sprow,spcol,rank);
4058: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4059: }
1.74 noro 4060: return r0;
4061: }
4062:
4063: FILE *nd_write,*nd_read;
4064:
4065: void nd_send_int(int a) {
4066: write_int(nd_write,&a);
4067: }
4068:
4069: void nd_send_intarray(int *p,int len) {
4070: write_intarray(nd_write,p,len);
4071: }
4072:
4073: int nd_recv_int() {
4074: int a;
4075:
4076: read_int(nd_read,&a);
4077: return a;
4078: }
4079:
4080: void nd_recv_intarray(int *p,int len) {
4081: read_intarray(nd_read,p,len);
4082: }
4083:
4084: void nd_send_ndv(NDV p) {
4085: int len,i;
4086: NMV m;
4087:
4088: if ( !p ) nd_send_int(0);
4089: else {
4090: len = LEN(p);
4091: nd_send_int(len);
4092: m = BDY(p);
4093: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4094: nd_send_int(CM(m));
4095: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4096: }
1.74 noro 4097: }
4098: }
4099:
4100: void nd_send_nd(ND p) {
4101: int len,i;
4102: NM m;
4103:
4104: if ( !p ) nd_send_int(0);
4105: else {
4106: len = LEN(p);
4107: nd_send_int(len);
4108: m = BDY(p);
4109: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4110: nd_send_int(CM(m));
4111: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4112: }
1.74 noro 4113: }
4114: }
1.65 noro 4115:
1.74 noro 4116: NDV nd_recv_ndv()
4117: {
4118: int len,i;
4119: NMV m,m0;
4120: NDV r;
1.65 noro 4121:
1.74 noro 4122: len = nd_recv_int();
4123: if ( !len ) return 0;
4124: else {
4125: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4126: ndv_alloc += len*nmv_adv;
4127: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4128: CM(m) = nd_recv_int();
4129: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4130: }
1.74 noro 4131: MKNDV(nd_nvar,m0,len,r);
4132: return r;
4133: }
4134: }
1.65 noro 4135:
1.74 noro 4136: int ox_exec_f4_red(Q proc)
4137: {
4138: Obj obj;
4139: STRING fname;
4140: NODE arg;
4141: int s;
4142: extern int ox_need_conv,ox_file_io;
4143:
4144: MKSTR(fname,"nd_exec_f4_red");
4145: arg = mknode(2,proc,fname);
4146: Pox_cmo_rpc(arg,&obj);
4147: s = get_ox_server_id(QTOS(proc));
4148: nd_write = iofp[s].out;
4149: nd_read = iofp[s].in;
4150: ox_need_conv = ox_file_io = 0;
4151: return s;
4152: }
4153:
4154: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4155: {
4156: int nsp,nred;
4157: int i,rank,s;
4158: NODE rp,r0,r;
4159: ND_pairs sp;
4160: NM_ind_pair pair;
4161: NMV nmv;
4162: NM nm;
4163: NDV nf;
4164: Obj proc,dmy;
4165:
4166: ox_launch_main(0,0,&proc);
4167: s = ox_exec_f4_red((Q)proc);
4168:
4169: nd_send_int(m);
4170: nd_send_int(nd_nvar);
4171: nd_send_int(nd_bpe);
4172: nd_send_int(nd_wpd);
4173: nd_send_int(nmv_adv);
4174:
1.89 noro 4175: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4176:
4177: nd_send_int(nd_psn);
4178: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4179:
4180: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4181: nd_send_int(nsp);
4182: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4183: nd_send_int(sp->i1); nd_send_int(sp->i2);
4184: }
4185:
4186: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4187:
4188: nred = length(rp0); nd_send_int(nred);
4189: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4190: pair = (NM_ind_pair)BDY(rp);
4191: nd_send_int(pair->index);
4192: nd_send_intarray(pair->mul->dl,nd_wpd);
4193: }
4194: fflush(nd_write);
4195: rank = nd_recv_int();
4196: fprintf(asir_out,"rank=%d\n",rank);
4197: r0 = 0;
4198: for ( i = 0; i < rank; i++ ) {
4199: nf = nd_recv_ndv();
4200: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4201: }
4202: Pox_shutdown(mknode(1,proc),&dmy);
4203: return r0;
4204: }
4205:
4206: /* server side */
4207:
4208: void nd_exec_f4_red_dist()
4209: {
4210: int m,i,nsp,col,s0size,nred,spcol,j,k;
4211: NM_ind_pair *rp0;
4212: NDV nf;
4213: UINT *s0vect;
4214: IndArray *imat;
4215: int *rhead;
4216: int **spmat;
4217: UINT *svect,*v;
4218: ND_pairs *sp0;
4219: int *colstat;
4220: int a,sprow,rank;
1.89 noro 4221: struct order_spec *ord;
1.74 noro 4222: Obj ordspec;
4223: ND spol;
1.76 noro 4224: int maxrs;
4225: int *spsugar;
1.74 noro 4226:
4227: nd_read = iofp[0].in;
4228: nd_write = iofp[0].out;
4229: m = nd_recv_int();
4230: nd_nvar = nd_recv_int();
4231: nd_bpe = nd_recv_int();
4232: nd_wpd = nd_recv_int();
4233: nmv_adv = nd_recv_int();
4234:
4235: loadobj(nd_read,&ordspec);
1.89 noro 4236: create_order_spec(0,ordspec,&ord);
4237: nd_init_ord(ord);
1.74 noro 4238: nd_setup_parameters(nd_nvar,0);
4239:
4240: nd_psn = nd_recv_int();
4241: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4242: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4243: for ( i = 0; i < nd_psn; i++ ) {
4244: nd_ps[i] = nd_recv_ndv();
4245: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4246: }
4247:
4248: nsp = nd_recv_int();
4249: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4250: for ( i = 0; i < nsp; i++ ) {
4251: NEWND_pairs(sp0[i]);
4252: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4253: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4254: }
4255:
4256: col = nd_recv_int();
4257: s0size = col*nd_wpd;
4258: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4259: nd_recv_intarray(s0vect,s0size);
4260:
4261: nred = nd_recv_int();
4262: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4263: for ( i = 0; i < nred; i++ ) {
4264: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4265: rp0[i]->index = nd_recv_int();
4266: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4267: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4268: }
4269:
4270: spcol = col-nred;
4271: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4272: rhead = (int *)MALLOC(col*sizeof(int));
4273: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4274:
4275: /* construction of index arrays */
4276: for ( i = 0; i < nred; i++ ) {
4277: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4278: rhead[imat[i]->head] = 1;
4279: }
4280:
4281: /* elimination (1st step) */
4282: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4283: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4284: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4285: for ( a = sprow = 0; a < nsp; a++ ) {
4286: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4287: if ( !spol ) continue;
1.74 noro 4288: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4289: if ( m == -1 )
4290: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4291: else
4292: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4293: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4294: if ( i < col ) {
4295: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4296: for ( j = k = 0; j < col; j++ )
4297: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4298: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4299: sprow++;
4300: }
1.76 noro 4301: nd_free(spol);
1.74 noro 4302: }
4303: /* elimination (2nd step) */
4304: colstat = (int *)ALLOCA(spcol*sizeof(int));
4305: if ( m == -1 )
1.76 noro 4306: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4307: else
1.76 noro 4308: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4309: nd_send_int(rank);
4310: for ( i = 0; i < rank; i++ ) {
4311: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4312: nd_send_ndv(nf);
1.63 noro 4313: }
1.74 noro 4314: fflush(nd_write);
1.76 noro 4315: }
4316:
4317: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4318: {
4319: int i,j,k,l,inv,a,rank,s;
4320: unsigned int *t,*pivot,*pk;
4321: unsigned int **mat;
4322:
4323: mat = (unsigned int **)mat0;
4324: for ( rank = 0, j = 0; j < col; j++ ) {
4325: for ( i = rank; i < row; i++ )
4326: mat[i][j] %= md;
4327: for ( i = rank; i < row; i++ )
4328: if ( mat[i][j] )
4329: break;
4330: if ( i == row ) {
4331: colstat[j] = 0;
4332: continue;
4333: } else
4334: colstat[j] = 1;
4335: if ( i != rank ) {
4336: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4337: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4338: }
4339: pivot = mat[rank];
4340: s = sugar[rank];
4341: inv = invm(pivot[j],md);
4342: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4343: if ( *pk ) {
4344: if ( *pk >= (unsigned int)md )
4345: *pk %= md;
4346: DMAR(*pk,inv,0,md,*pk)
4347: }
4348: for ( i = rank+1; i < row; i++ ) {
4349: t = mat[i];
4350: if ( a = t[j] ) {
4351: sugar[i] = MAX(sugar[i],s);
4352: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4353: }
4354: }
4355: rank++;
4356: }
4357: for ( j = col-1, l = rank-1; j >= 0; j-- )
4358: if ( colstat[j] ) {
4359: pivot = mat[l];
4360: s = sugar[l];
4361: for ( i = 0; i < l; i++ ) {
4362: t = mat[i];
4363: t[j] %= md;
4364: if ( a = t[j] ) {
4365: sugar[i] = MAX(sugar[i],s);
4366: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4367: }
4368: }
4369: l--;
4370: }
4371: for ( j = 0, l = 0; l < rank; j++ )
4372: if ( colstat[j] ) {
4373: t = mat[l];
4374: for ( k = j; k < col; k++ )
4375: if ( t[k] >= (unsigned int)md )
4376: t[k] %= md;
4377: l++;
4378: }
4379: return rank;
4380: }
4381:
4382: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4383: {
4384: int i,j,k,l,inv,a,rank,s;
4385: unsigned int *t,*pivot,*pk;
4386: unsigned int **mat;
4387:
4388: mat = (unsigned int **)mat0;
4389: for ( rank = 0, j = 0; j < col; j++ ) {
4390: for ( i = rank; i < row; i++ )
4391: if ( mat[i][j] )
4392: break;
4393: if ( i == row ) {
4394: colstat[j] = 0;
4395: continue;
4396: } else
4397: colstat[j] = 1;
4398: if ( i != rank ) {
4399: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4400: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4401: }
4402: pivot = mat[rank];
4403: s = sugar[rank];
4404: inv = _invsf(pivot[j]);
4405: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4406: if ( *pk )
4407: *pk = _mulsf(*pk,inv);
4408: for ( i = rank+1; i < row; i++ ) {
4409: t = mat[i];
4410: if ( a = t[j] ) {
4411: sugar[i] = MAX(sugar[i],s);
4412: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4413: }
4414: }
4415: rank++;
4416: }
4417: for ( j = col-1, l = rank-1; j >= 0; j-- )
4418: if ( colstat[j] ) {
4419: pivot = mat[l];
4420: s = sugar[l];
4421: for ( i = 0; i < l; i++ ) {
4422: t = mat[i];
4423: if ( a = t[j] ) {
4424: sugar[i] = MAX(sugar[i],s);
4425: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4426: }
4427: }
4428: l--;
4429: }
4430: return rank;
1.77 noro 4431: }
4432:
4433: int ndv_ishomo(NDV p)
4434: {
4435: NMV m;
4436: int len,h;
4437:
4438: if ( !p ) return 1;
4439: len = LEN(p);
4440: m = BDY(p);
4441: h = TD(DL(m));
4442: NMV_ADV(m);
4443: for ( len--; len; len--, NMV_ADV(m) )
4444: if ( TD(DL(m)) != h ) return 0;
4445: return 1;
4446: }
4447:
4448: void ndv_save(NDV p,int index)
4449: {
4450: FILE *s;
4451: char name[BUFSIZ];
4452: short id;
4453: int nv,sugar,len,n,i,td,e,j;
4454: NMV m;
4455: unsigned int *dl;
4456:
4457: sprintf(name,"%s/%d",Demand,index);
4458: s = fopen(name,"w");
4459: savevl(s,0);
4460: if ( !p ) {
4461: saveobj(s,0);
4462: return;
4463: }
4464: id = O_DP;
4465: nv = NV(p);
4466: sugar = SG(p);
4467: len = LEN(p);
4468: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
4469: write_int(s,&len);
4470:
4471: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4472: saveobj(s,(Obj)CQ(m));
4473: dl = DL(m);
4474: td = TD(dl);
4475: write_int(s,&td);
4476: for ( j = 0; j < nv; j++ ) {
4477: e = GET_EXP(dl,j);
4478: write_int(s,&e);
4479: }
4480: }
4481: fclose(s);
4482: }
4483:
4484: NDV ndv_load(int index)
4485: {
4486: FILE *s;
4487: char name[BUFSIZ];
4488: short id;
4489: int nv,sugar,len,n,i,td,e,j;
4490: NDV d;
4491: NMV m0,m;
4492: unsigned int *dl;
4493: Obj obj;
4494:
4495: sprintf(name,"%s/%d",Demand,index);
4496: s = fopen(name,"r");
4497: if ( !s ) return 0;
4498:
4499: skipvl(s);
4500: read_short(s,&id);
4501: if ( !id ) return 0;
4502: read_int(s,&nv);
4503: read_int(s,&sugar);
4504: read_int(s,&len);
4505:
4506: m0 = m = MALLOC(len*nmv_adv);
4507: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4508: loadobj(s,&obj); CQ(m) = (Q)obj;
4509: dl = DL(m);
4510: ndl_zero(dl);
4511: read_int(s,&td); TD(dl) = td;
4512: for ( j = 0; j < nv; j++ ) {
4513: read_int(s,&e);
4514: PUT_EXP(dl,j,e);
4515: }
4516: if ( nd_blockmask ) ndl_weight_mask(dl);
4517: }
4518: fclose(s);
4519: MKNDV(nv,m0,len,d);
4520: SG(d) = sugar;
4521: return d;
1.59 noro 4522: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>