![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to nd.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [local] / OpenXM_contrib2 / asir2000 / engine|
Return to nd.c CVS log | Up to [local] / OpenXM_contrib2 / asir2000 / engine |
1.95 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.94 2004/03/15 07:30:44 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;
1.20 noro 32:
1.42 noro 33: static RHist *nd_red;
1.1 noro 34:
1.42 noro 35: static int nd_found,nd_create,nd_notfirst;
36: static int nmv_adv;
1.77 noro 37: static int nd_demand;
1.1 noro 38:
39: void nd_free_private_storage()
40: {
41: _nm_free_list = 0;
1.5 noro 42: _ndp_free_list = 0;
1.71 noro 43: #if 0
1.1 noro 44: GC_gcollect();
1.71 noro 45: #endif
1.1 noro 46: }
47:
48: void _NM_alloc()
49: {
50: NM p;
51: int i;
52:
1.11 noro 53: for ( i = 0; i < 1024; i++ ) {
1.61 noro 54: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 55: p->next = _nm_free_list; _nm_free_list = p;
56: }
57: }
58:
59: void _ND_alloc()
60: {
61: ND p;
62: int i;
63:
64: for ( i = 0; i < 1024; i++ ) {
65: p = (ND)GC_malloc(sizeof(struct oND));
66: p->body = (NM)_nd_free_list; _nd_free_list = p;
67: }
68: }
69:
70: void _NDP_alloc()
71: {
72: ND_pairs p;
73: int i;
74:
1.11 noro 75: for ( i = 0; i < 1024; i++ ) {
1.1 noro 76: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 77: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 78: p->next = _ndp_free_list; _ndp_free_list = p;
79: }
80: }
81:
1.30 noro 82: INLINE int nd_length(ND p)
1.1 noro 83: {
84: NM m;
85: int i;
86:
87: if ( !p )
88: return 0;
89: else {
90: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
91: return i;
92: }
93: }
94:
1.61 noro 95: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 96: {
1.61 noro 97: UINT u1,u2;
1.1 noro 98: int i,j;
99:
1.34 noro 100: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 101: #if USE_UNROLL
1.1 noro 102: switch ( nd_bpe ) {
1.62 noro 103: case 3:
104: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
105: u1 = d1[i]; u2 = d2[i];
106: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
107: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
108: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
109: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
110: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
111: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
112: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
113: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
114: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
115: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
116: }
117: return 1;
118: break;
1.1 noro 119: case 4:
1.41 noro 120: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 121: u1 = d1[i]; u2 = d2[i];
122: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 123: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
124: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
125: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
126: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
127: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
128: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
129: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 130: }
131: return 1;
132: break;
133: case 6:
1.41 noro 134: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 135: u1 = d1[i]; u2 = d2[i];
136: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 137: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
138: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
139: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
140: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 141: }
142: return 1;
143: break;
144: case 8:
1.41 noro 145: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 146: u1 = d1[i]; u2 = d2[i];
147: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 148: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
149: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
150: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 151: }
152: return 1;
153: break;
154: case 16:
1.41 noro 155: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 156: u1 = d1[i]; u2 = d2[i];
157: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 158: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 159: }
160: return 1;
161: break;
162: case 32:
1.41 noro 163: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 164: if ( d1[i] < d2[i] ) return 0;
165: return 1;
166: break;
167: default:
1.41 noro 168: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 169: u1 = d1[i]; u2 = d2[i];
170: for ( j = 0; j < nd_epw; j++ )
171: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
172: }
173: return 1;
174: }
1.65 noro 175: #else
176: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
177: u1 = d1[i]; u2 = d2[i];
178: for ( j = 0; j < nd_epw; j++ )
179: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
180: }
181: return 1;
182: #endif
1.1 noro 183: }
184:
1.61 noro 185: /*
186: * If the current order is a block order,
187: * then the last block is length 1 and contains
188: * the homo variable. Otherwise, the original
189: * order is either 0 or 2.
190: */
191:
192: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 193: {
1.61 noro 194: int w,i,e,n,omask0;
195:
196: omask0 = (1<<obpe)-1;
197: n = nd_nvar-1;
198: ndl_zero(r);
199: for ( i = 0; i < n; i++ ) {
200: e = GET_EXP_OLD(d,i);
201: PUT_EXP(r,i,e);
202: }
203: w = TD(d);
204: PUT_EXP(r,nd_nvar-1,weight-w);
205: TD(r) = weight;
206: if ( nd_blockmask ) ndl_weight_mask(r);
207: }
208:
209: void ndl_dehomogenize(UINT *d)
210: {
211: UINT mask;
212: UINT h;
1.31 noro 213: int i,bits;
1.23 noro 214:
1.44 noro 215: if ( nd_blockmask ) {
216: h = GET_EXP(d,nd_nvar-1);
1.45 noro 217: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 218: TD(d) -= h;
219: d[nd_exporigin-1] -= h;
220: } else {
221: if ( nd_isrlex ) {
222: if ( nd_bpe == 32 ) {
223: h = d[nd_exporigin];
224: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
225: d[i-1] = d[i];
226: d[i-1] = 0;
227: TD(d) -= h;
228: } else {
229: bits = nd_epw*nd_bpe;
230: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
231: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
232: for ( i = nd_exporigin; i < nd_wpd; i++ )
233: d[i] = ((d[i]<<nd_bpe)&mask)
234: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
235: TD(d) -= h;
236: }
1.45 noro 237: } else {
238: h = GET_EXP(d,nd_nvar-1);
239: XOR_EXP(d,nd_nvar-1,h);
240: TD(d) -= h;
241: }
1.44 noro 242: }
1.23 noro 243: }
244:
1.61 noro 245: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 246: {
1.61 noro 247: UINT t1,t2,u,u1,u2;
1.43 noro 248: int i,j,l;
1.1 noro 249:
1.65 noro 250: #if USE_UNROLL
1.1 noro 251: switch ( nd_bpe ) {
1.62 noro 252: case 3:
253: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
254: u1 = d1[i]; u2 = d2[i];
255: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
256: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
257: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
258: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
259: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
260: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
261: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
262: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
263: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
264: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
265: d[i] = u;
266: }
267: break;
1.1 noro 268: case 4:
1.41 noro 269: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 270: u1 = d1[i]; u2 = d2[i];
271: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 272: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
273: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
274: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
275: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
276: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
277: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
278: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 279: d[i] = u;
280: }
281: break;
282: case 6:
1.41 noro 283: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 284: u1 = d1[i]; u2 = d2[i];
285: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 286: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
287: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
288: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
289: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 290: d[i] = u;
291: }
292: break;
293: case 8:
1.41 noro 294: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 295: u1 = d1[i]; u2 = d2[i];
296: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 297: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
298: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
299: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 300: d[i] = u;
301: }
302: break;
303: case 16:
1.41 noro 304: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 305: u1 = d1[i]; u2 = d2[i];
306: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 307: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 308: d[i] = u;
309: }
310: break;
311: case 32:
1.41 noro 312: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 313: u1 = d1[i]; u2 = d2[i];
314: d[i] = u1>u2?u1:u2;
315: }
316: break;
317: default:
1.41 noro 318: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 319: u1 = d1[i]; u2 = d2[i];
320: for ( j = 0, u = 0; j < nd_epw; j++ ) {
321: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
322: }
323: d[i] = u;
324: }
325: break;
326: }
1.65 noro 327: #else
328: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
329: u1 = d1[i]; u2 = d2[i];
330: for ( j = 0, u = 0; j < nd_epw; j++ ) {
331: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
332: }
333: d[i] = u;
334: }
335: #endif
1.39 noro 336: TD(d) = ndl_weight(d);
1.61 noro 337: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 338: }
339:
1.61 noro 340: int ndl_weight(UINT *d)
1.1 noro 341: {
1.61 noro 342: UINT t,u;
1.1 noro 343: int i,j;
344:
1.60 noro 345: if ( current_dl_weight_vector )
346: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
347: u = GET_EXP(d,i);
348: t += MUL_WEIGHT(u,i);
349: }
350: else
351: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
352: u = d[i];
353: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
354: t += (u&nd_mask0);
355: }
1.1 noro 356: return t;
357: }
358:
1.61 noro 359: void ndl_weight_mask(UINT *d)
1.43 noro 360: {
1.61 noro 361: UINT t,u;
362: UINT *mask;
363: int i,j,k,l;
1.43 noro 364:
1.61 noro 365: l = nd_blockmask->n;
366: for ( k = 0; k < l; k++ ) {
367: mask = nd_blockmask->mask[k];
368: if ( current_dl_weight_vector )
369: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
370: u = GET_EXP_MASK(d,i,mask);
371: t += MUL_WEIGHT(u,i);
372: }
373: else
374: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
375: u = d[i]&mask[i];
376: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
377: t += (u&nd_mask0);
378: }
379: d[k+1] = t;
380: }
1.43 noro 381: }
382:
1.61 noro 383: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 384: {
385: int i;
386:
1.41 noro 387: d1 += nd_exporigin;
388: d2 += nd_exporigin;
389: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 390: if ( *d1 > *d2 )
1.32 noro 391: return nd_isrlex ? -1 : 1;
1.1 noro 392: else if ( *d1 < *d2 )
1.32 noro 393: return nd_isrlex ? 1 : -1;
1.1 noro 394: return 0;
395: }
396:
1.61 noro 397: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 398: {
399: int i,l,j,ord_o,ord_l;
400: struct order_pair *op;
1.61 noro 401: UINT t1,t2,m;
402: UINT *mask;
1.43 noro 403:
404: l = nd_blockmask->n;
405: op = nd_blockmask->order_pair;
406: for ( j = 0; j < l; j++ ) {
407: mask = nd_blockmask->mask[j];
408: ord_o = op[j].order;
409: if ( ord_o < 2 )
1.44 noro 410: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
411: else if ( t1 < t2 ) return -1;
1.43 noro 412: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 413: m = mask[i];
414: t1 = d1[i]&m;
415: t2 = d2[i]&m;
1.43 noro 416: if ( t1 > t2 )
417: return !ord_o ? -1 : 1;
418: else if ( t1 < t2 )
419: return !ord_o ? 1 : -1;
420: }
421: }
422: return 0;
423: }
424:
1.58 noro 425: /* TDH -> WW -> TD-> RL */
426:
1.61 noro 427: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 428: {
429: int i,m,e1,e2;
430:
431: if ( TD(d1) > TD(d2) ) return 1;
432: else if ( TD(d1) < TD(d2) ) return -1;
433: m = nd_nvar>>1;
434: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
435: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
436: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
437: }
438: if ( e1 > e2 ) return 1;
439: else if ( e1 < e2 ) return -1;
440: return ndl_lex_compare(d1,d2);
441: }
442:
1.61 noro 443: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 444: {
445: int i;
446:
1.81 noro 447: switch ( nd_wpd ) {
448: case 2:
449: if ( TD(d2) != TD(d1) ) return 0;
450: if ( d2[1] != d1[1] ) return 0;
451: return 1;
452: break;
453: case 3:
454: if ( TD(d2) != TD(d1) ) return 0;
455: if ( d2[1] != d1[1] ) return 0;
456: if ( d2[2] != d1[2] ) return 0;
457: return 1;
458: break;
459: default:
460: for ( i = 0; i < nd_wpd; i++ )
461: if ( *d1++ != *d2++ ) return 0;
462: return 1;
463: break;
464: }
1.1 noro 465: }
466:
1.61 noro 467: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 468: {
469: int i;
470:
471: switch ( nd_wpd ) {
1.41 noro 472: case 2:
1.34 noro 473: TD(d2) = TD(d1);
474: d2[1] = d1[1];
1.6 noro 475: break;
1.41 noro 476: case 3:
1.34 noro 477: TD(d2) = TD(d1);
1.6 noro 478: d2[1] = d1[1];
1.34 noro 479: d2[2] = d1[2];
1.6 noro 480: break;
481: default:
1.41 noro 482: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 483: d2[i] = d1[i];
484: break;
485: }
486: }
487:
1.61 noro 488: INLINE void ndl_zero(UINT *d)
489: {
490: int i;
491: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
492: }
493:
494: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 495: {
496: int i;
497:
1.43 noro 498: #if 1
1.6 noro 499: switch ( nd_wpd ) {
1.41 noro 500: case 2:
501: TD(d) = TD(d1)+TD(d2);
1.34 noro 502: d[1] = d1[1]+d2[1];
1.6 noro 503: break;
1.41 noro 504: case 3:
505: TD(d) = TD(d1)+TD(d2);
1.6 noro 506: d[1] = d1[1]+d2[1];
1.34 noro 507: d[2] = d1[2]+d2[2];
1.6 noro 508: break;
509: default:
1.43 noro 510: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 511: break;
512: }
1.43 noro 513: #else
514: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
515: #endif
1.6 noro 516: }
517:
1.55 noro 518: /* d1 += d2 */
1.61 noro 519: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 520: {
521: int i;
522:
523: #if 1
524: switch ( nd_wpd ) {
525: case 2:
526: TD(d1) += TD(d2);
527: d1[1] += d2[1];
528: break;
529: case 3:
530: TD(d1) += TD(d2);
531: d1[1] += d2[1];
532: d1[2] += d2[2];
533: break;
534: default:
535: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
536: break;
537: }
538: #else
539: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
540: #endif
541: }
542:
1.61 noro 543: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 544: {
545: int i;
546:
1.43 noro 547: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 548: }
549:
1.61 noro 550: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 551: {
1.61 noro 552: UINT t1,t2,u,u1,u2;
1.1 noro 553: int i,j;
554:
1.65 noro 555: #if USE_UNROLL
1.1 noro 556: switch ( nd_bpe ) {
1.62 noro 557: case 3:
558: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
559: u1 = d1[i]; u2 = d2[i];
560: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
561: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
562: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
563: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
564: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
565: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
566: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
567: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
568: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
569: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
570: }
571: return 1;
572: break;
1.1 noro 573: case 4:
1.41 noro 574: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 575: u1 = d1[i]; u2 = d2[i];
576: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 577: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
578: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
579: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
580: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
581: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
582: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
583: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 584: }
585: return 1;
586: break;
587: case 6:
1.41 noro 588: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 589: u1 = d1[i]; u2 = d2[i];
590: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 591: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
592: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
593: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
594: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 595: }
596: return 1;
597: break;
598: case 8:
1.41 noro 599: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 600: u1 = d1[i]; u2 = d2[i];
601: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 602: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
603: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
604: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 605: }
606: return 1;
607: break;
608: case 16:
1.41 noro 609: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 610: u1 = d1[i]; u2 = d2[i];
611: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 612: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 613: }
614: return 1;
615: break;
616: case 32:
1.41 noro 617: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 618: if ( d1[i] && d2[i] ) return 0;
619: return 1;
620: break;
621: default:
1.41 noro 622: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 623: u1 = d1[i]; u2 = d2[i];
624: for ( j = 0; j < nd_epw; j++ ) {
625: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
626: u1 >>= nd_bpe; u2 >>= nd_bpe;
627: }
628: }
629: return 1;
630: break;
631: }
1.65 noro 632: #else
633: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
634: u1 = d1[i]; u2 = d2[i];
635: for ( j = 0; j < nd_epw; j++ ) {
636: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
637: u1 >>= nd_bpe; u2 >>= nd_bpe;
638: }
639: }
640: return 1;
641: #endif
1.1 noro 642: }
643:
1.61 noro 644: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 645: {
1.61 noro 646: UINT u2;
647: UINT *d1;
1.5 noro 648: int i,j,ind,k;
1.1 noro 649:
1.5 noro 650: d1 = nd_bound[index];
651: ind = 0;
1.65 noro 652: #if USE_UNROLL
1.5 noro 653: switch ( nd_bpe ) {
1.62 noro 654: case 3:
655: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
656: u2 = d2[i];
657: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
658: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
659: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
660: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
661: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
662: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
663: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
664: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
665: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
666: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
667: }
668: return 0;
669: break;
1.5 noro 670: case 4:
1.41 noro 671: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 672: u2 = d2[i];
673: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
674: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
675: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
676: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
677: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
678: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
679: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
680: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
681: }
682: return 0;
683: break;
684: case 6:
1.41 noro 685: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 686: u2 = d2[i];
687: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
688: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
689: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
690: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
691: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
692: }
693: return 0;
694: break;
695: case 8:
1.41 noro 696: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 697: u2 = d2[i];
698: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
699: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
700: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
701: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
702: }
703: return 0;
704: break;
705: case 16:
1.41 noro 706: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 707: u2 = d2[i];
708: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
709: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
710: }
711: return 0;
712: break;
713: case 32:
1.41 noro 714: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 715: if ( d1[i]+d2[i]<d1[i] ) return 1;
716: return 0;
717: break;
718: default:
1.41 noro 719: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 720: u2 = d2[i];
721: k = (nd_epw-1)*nd_bpe;
722: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
723: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
724: }
725: return 0;
726: break;
727: }
1.65 noro 728: #else
729: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
730: u2 = d2[i];
731: k = (nd_epw-1)*nd_bpe;
732: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
733: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
734: }
735: return 0;
736: #endif
1.1 noro 737: }
738:
1.61 noro 739: INLINE int ndl_hash_value(UINT *d)
1.1 noro 740: {
741: int i;
742: int r;
743:
1.34 noro 744: r = 0;
1.41 noro 745: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 746: r = ((r<<16)+d[i])%REDTAB_LEN;
747: return r;
748: }
749:
1.63 noro 750: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 751: {
1.13 noro 752: RHist r;
1.6 noro 753: int d,k,i;
1.1 noro 754:
1.63 noro 755: d = ndl_hash_value(dg);
1.13 noro 756: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 757: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 758: if ( k > 0 ) nd_notfirst++;
759: nd_found++;
1.13 noro 760: return r->index;
1.1 noro 761: }
762: }
1.13 noro 763: if ( Reverse )
764: for ( i = nd_psn-1; i >= 0; i-- ) {
765: r = nd_psh[i];
1.34 noro 766: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 767: nd_create++;
1.34 noro 768: nd_append_red(dg,i);
1.13 noro 769: return i;
770: }
771: }
772: else
773: for ( i = 0; i < nd_psn; i++ ) {
774: r = nd_psh[i];
1.34 noro 775: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 776: nd_create++;
1.34 noro 777: nd_append_red(dg,i);
1.13 noro 778: return i;
779: }
1.1 noro 780: }
1.6 noro 781: return -1;
1.1 noro 782: }
783:
1.63 noro 784: ND nd_merge(ND p1,ND p2)
785: {
786: int n,c;
787: int t,can,td1,td2;
788: ND r;
789: NM m1,m2,mr0,mr,s;
790:
791: if ( !p1 ) return p2;
792: else if ( !p2 ) return p1;
793: else {
794: can = 0;
795: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
796: c = DL_COMPARE(DL(m1),DL(m2));
797: switch ( c ) {
798: case 0:
799: s = m1; m1 = NEXT(m1);
800: can++; NEXTNM2(mr0,mr,s);
801: s = m2; m2 = NEXT(m2); FREENM(s);
802: break;
803: case 1:
804: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
805: break;
806: case -1:
807: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
808: break;
809: }
810: }
811: if ( !mr0 )
812: if ( m1 ) mr0 = m1;
813: else if ( m2 ) mr0 = m2;
814: else return 0;
815: else if ( m1 ) NEXT(mr) = m1;
816: else if ( m2 ) NEXT(mr) = m2;
817: else NEXT(mr) = 0;
818: BDY(p1) = mr0;
819: SG(p1) = MAX(SG(p1),SG(p2));
820: LEN(p1) = LEN(p1)+LEN(p2)-can;
821: FREEND(p2);
822: return p1;
823: }
824: }
825:
1.31 noro 826: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 827: {
828: int n,c;
1.34 noro 829: int t,can,td1,td2;
1.1 noro 830: ND r;
831: NM m1,m2,mr0,mr,s;
832:
1.34 noro 833: if ( !p1 ) return p2;
834: else if ( !p2 ) return p1;
1.71 noro 835: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 836: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 837: else {
1.30 noro 838: can = 0;
1.1 noro 839: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 840: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 841: switch ( c ) {
842: case 0:
1.19 noro 843: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 844: if ( t < 0 ) t += mod;
1.1 noro 845: s = m1; m1 = NEXT(m1);
846: if ( t ) {
1.34 noro 847: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.17 noro 848: } else {
1.34 noro 849: can += 2; FREENM(s);
1.17 noro 850: }
851: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 ! noro 852: break;
! 853: case 1:
! 854: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
! 855: break;
! 856: case -1:
! 857: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
! 858: break;
! 859: }
! 860: }
! 861: if ( !mr0 )
! 862: if ( m1 ) mr0 = m1;
! 863: else if ( m2 ) mr0 = m2;
! 864: else return 0;
! 865: else if ( m1 ) NEXT(mr) = m1;
! 866: else if ( m2 ) NEXT(mr) = m2;
! 867: else NEXT(mr) = 0;
! 868: BDY(p1) = mr0;
! 869: SG(p1) = MAX(SG(p1),SG(p2));
! 870: LEN(p1) = LEN(p1)+LEN(p2)-can;
! 871: FREEND(p2);
! 872: return p1;
! 873: }
! 874: }
! 875:
! 876: /* XXX on opteron, the inlined manipulation of destructive additon of
! 877: * two NM seems to make gcc optimizer get confused, so the part is
! 878: * done in a function.
! 879: */
! 880:
! 881: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
! 882: {
! 883: NM s;
! 884: Q t;
! 885: int can;
! 886:
! 887: addq(CQ(*m1),CQ(*m2),&t);
! 888: s = *m1; *m1 = NEXT(*m1);
! 889: if ( t ) {
! 890: can = 1; NEXTNM2(*mr0,*mr,s); CQ(*mr) = (t);
! 891: } else {
! 892: can = 2; FREENM(s);
! 893: }
! 894: s = *m2; *m2 = NEXT(*m2); FREENM(s);
! 895: return can;
! 896: }
! 897:
! 898: ND nd_add_q(ND p1,ND p2)
! 899: {
! 900: int n,c,can;
! 901: ND r;
! 902: NM m1,m2,mr0,mr,s;
! 903: Q t;
! 904:
! 905: if ( !p1 ) return p2;
! 906: else if ( !p2 ) return p1;
! 907: else {
! 908: can = 0;
! 909: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
! 910: c = DL_COMPARE(DL(m1),DL(m2));
! 911: switch ( c ) {
! 912: case 0:
! 913: #if defined(__x86_64__)
! 914: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
! 915: #else
! 916: addq(CQ(m1),CQ(m2),&t);
! 917: s = m1; m1 = NEXT(m1);
! 918: if ( t ) {
! 919: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
! 920: } else {
! 921: can += 2; FREENM(s);
! 922: }
! 923: s = m2; m2 = NEXT(m2); FREENM(s);
! 924: #endif
1.17 noro 925: break;
926: case 1:
927: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
928: break;
929: case -1:
930: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
931: break;
932: }
933: }
934: if ( !mr0 )
1.34 noro 935: if ( m1 ) mr0 = m1;
936: else if ( m2 ) mr0 = m2;
937: else return 0;
938: else if ( m1 ) NEXT(mr) = m1;
939: else if ( m2 ) NEXT(mr) = m2;
940: else NEXT(mr) = 0;
1.17 noro 941: BDY(p1) = mr0;
942: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 943: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 944: FREEND(p2);
945: return p1;
946: }
947: }
948:
1.71 noro 949: ND nd_add_sf(ND p1,ND p2)
950: {
951: int n,c,can;
952: ND r;
953: NM m1,m2,mr0,mr,s;
954: int t;
955:
956: if ( !p1 ) return p2;
957: else if ( !p2 ) return p1;
958: else {
959: can = 0;
960: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
961: c = DL_COMPARE(DL(m1),DL(m2));
962: switch ( c ) {
963: case 0:
964: t = _addsf(CM(m1),CM(m2));
965: s = m1; m1 = NEXT(m1);
966: if ( t ) {
967: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
968: } else {
969: can += 2; FREENM(s);
970: }
971: s = m2; m2 = NEXT(m2); FREENM(s);
972: break;
973: case 1:
974: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
975: break;
976: case -1:
977: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
978: break;
979: }
980: }
981: if ( !mr0 )
982: if ( m1 ) mr0 = m1;
983: else if ( m2 ) mr0 = m2;
984: else return 0;
985: else if ( m1 ) NEXT(mr) = m1;
986: else if ( m2 ) NEXT(mr) = m2;
987: else NEXT(mr) = 0;
988: BDY(p1) = mr0;
989: SG(p1) = MAX(SG(p1),SG(p2));
990: LEN(p1) = LEN(p1)+LEN(p2)-can;
991: FREEND(p2);
992: return p1;
993: }
994: }
995:
1.1 noro 996: /* ret=1 : success, ret=0 : overflow */
1.69 noro 997: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 998: {
1.11 noro 999: ND d;
1.1 noro 1000: NM m,mrd,tail;
1.7 noro 1001: NM mul;
1.10 noro 1002: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1003: int c,c1,c2,dummy;
1.17 noro 1004: RHist h;
1.11 noro 1005: NDV p,red;
1.69 noro 1006: Q cg,cred,gcd,tq,qq;
1.21 noro 1007: double hmag;
1.1 noro 1008:
1.69 noro 1009: if ( dn ) {
1010: if ( mod )
1011: dn->m = 1;
1012: else
1013: dn->z = ONE;
1014: }
1.1 noro 1015: if ( !g ) {
1016: *rp = 0;
1017: return 1;
1018: }
1.34 noro 1019: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1020:
1.14 noro 1021: sugar0 = sugar = SG(g);
1.1 noro 1022: n = NV(g);
1.61 noro 1023: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1024: for ( d = 0; g; ) {
1.63 noro 1025: index = ndl_find_reducer(HDL(g));
1.6 noro 1026: if ( index >= 0 ) {
1.17 noro 1027: h = nd_psh[index];
1028: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1029: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1030: nd_free(g); nd_free(d);
1031: return 0;
1032: }
1.77 noro 1033: if ( nd_demand )
1034: p = ndv_load(index);
1035: else
1036: p = ps[index];
1.71 noro 1037: if ( mod == -1 )
1038: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1039: else if ( mod ) {
1.19 noro 1040: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1041: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1042: } else {
1.17 noro 1043: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1044: chsgnq(cg,&CQ(mul));
1.20 noro 1045: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1046: if ( dn ) {
1047: mulq(dn->z,cred,&tq); dn->z = tq;
1048: }
1.16 noro 1049: }
1.55 noro 1050: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1051: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1052: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1053: tq = HCQ(g);
1.21 noro 1054: nd_removecont2(d,g);
1.69 noro 1055: if ( dn ) {
1056: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1057: }
1.21 noro 1058: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1059: }
1.1 noro 1060: } else if ( !full ) {
1061: *rp = g;
1062: return 1;
1063: } else {
1064: m = BDY(g);
1065: if ( NEXT(m) ) {
1.34 noro 1066: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1067: } else {
1068: FREEND(g); g = 0;
1069: }
1070: if ( d ) {
1.34 noro 1071: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1072: } else {
1.34 noro 1073: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1074: }
1075: }
1076: }
1.34 noro 1077: if ( d ) SG(d) = sugar;
1.1 noro 1078: *rp = d;
1079: return 1;
1080: }
1.28 noro 1081:
1.53 noro 1082: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1083: {
1084: int hindex,index;
1085: NDV p;
1086: ND u,d,red;
1087: NODE l;
1.31 noro 1088: NM mul,m,mrd,tail;
1.25 noro 1089: int sugar,psugar,n,h_reducible;
1090: PGeoBucket bucket;
1091: int c,c1,c2;
1.26 noro 1092: Q cg,cred,gcd,zzz;
1.25 noro 1093: RHist h;
1.28 noro 1094: double hmag,gmag;
1.77 noro 1095: int count = 0;
1096: int hcount = 0;
1.25 noro 1097:
1098: if ( !g ) {
1099: *rp = 0;
1100: return 1;
1101: }
1102: sugar = SG(g);
1103: n = NV(g);
1.34 noro 1104: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1105: bucket = create_pbucket();
1.31 noro 1106: add_pbucket(mod,bucket,g);
1.25 noro 1107: d = 0;
1.61 noro 1108: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1109: while ( 1 ) {
1.26 noro 1110: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1111: if ( hindex < 0 ) {
1.77 noro 1112: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1113: if ( d ) SG(d) = sugar;
1.25 noro 1114: *rp = d;
1115: return 1;
1116: }
1117: g = bucket->body[hindex];
1.63 noro 1118: index = ndl_find_reducer(HDL(g));
1.25 noro 1119: if ( index >= 0 ) {
1.77 noro 1120: count++;
1121: if ( !d ) hcount++;
1.25 noro 1122: h = nd_psh[index];
1123: ndl_sub(HDL(g),DL(h),DL(mul));
1124: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1125: nd_free(d);
1.25 noro 1126: free_pbucket(bucket);
1127: *rp = 0;
1128: return 0;
1129: }
1.53 noro 1130: p = ps[index];
1.71 noro 1131: if ( mod == -1 )
1132: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1133: else if ( mod ) {
1.25 noro 1134: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1135: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1136: } else {
1137: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1138: chsgnq(cg,&CQ(mul));
1.26 noro 1139: nd_mul_c_q(d,cred);
1140: mulq_pbucket(bucket,cred);
1141: g = bucket->body[hindex];
1.28 noro 1142: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1143: }
1.55 noro 1144: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1145: bucket->body[hindex] = nd_remove_head(g);
1146: red = nd_remove_head(red);
1.31 noro 1147: add_pbucket(mod,bucket,red);
1.34 noro 1148: psugar = SG(p)+TD(DL(mul));
1149: sugar = MAX(sugar,psugar);
1.28 noro 1150: if ( !mod && hmag && (gmag > hmag) ) {
1151: g = normalize_pbucket(mod,bucket);
1152: if ( !g ) {
1.34 noro 1153: if ( d ) SG(d) = sugar;
1.28 noro 1154: *rp = d;
1155: return 1;
1156: }
1157: nd_removecont2(d,g);
1158: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1159: add_pbucket(mod,bucket,g);
1.28 noro 1160: }
1.25 noro 1161: } else if ( !full ) {
1162: g = normalize_pbucket(mod,bucket);
1.34 noro 1163: if ( g ) SG(g) = sugar;
1.25 noro 1164: *rp = g;
1165: return 1;
1166: } else {
1167: m = BDY(g);
1168: if ( NEXT(m) ) {
1.34 noro 1169: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1170: } else {
1171: FREEND(g); g = 0;
1172: }
1173: bucket->body[hindex] = g;
1174: NEXT(m) = 0;
1175: if ( d ) {
1.34 noro 1176: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1177: } else {
1.34 noro 1178: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1179: }
1180: }
1181: }
1182: }
1.27 noro 1183:
1.61 noro 1184: /* input : list of NDV, cand : list of NDV */
1.28 noro 1185:
1.61 noro 1186: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1187: {
1188: int n,i,stat;
1189: ND nf,d;
1.61 noro 1190: NDV r;
1.45 noro 1191: NODE t,s;
1.69 noro 1192: union oNDC dn;
1.45 noro 1193:
1.61 noro 1194: ndv_setup(0,0,cand);
1.31 noro 1195: n = length(cand);
1.28 noro 1196:
1197: /* membercheck : list is a subset of Id(cand) ? */
1198: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1199: again:
1.61 noro 1200: if ( nd_bpe > obpe )
1201: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1202: else
1203: r = (NDV)BDY(t);
1204: d = ndvtond(0,r);
1.69 noro 1205: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1206: if ( !stat ) {
1207: nd_reconstruct(0,0,0);
1208: goto again;
1209: } else if ( nf ) return 0;
1.71 noro 1210: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1211: }
1.71 noro 1212: if ( DP_Print ) { printf("\n"); }
1.28 noro 1213: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1214: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1215: /* XXX */
1.23 noro 1216: return 1;
1217: }
1.1 noro 1218:
1219: ND nd_remove_head(ND p)
1220: {
1221: NM m;
1222:
1223: m = BDY(p);
1224: if ( !NEXT(m) ) {
1.34 noro 1225: FREEND(p); p = 0;
1.31 noro 1226: } else {
1.34 noro 1227: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1228: }
1.1 noro 1229: FREENM(m);
1230: return p;
1231: }
1232:
1.69 noro 1233: ND nd_separate_head(ND p,ND *head)
1234: {
1235: NM m,m0;
1236: ND r;
1237:
1238: m = BDY(p);
1239: if ( !NEXT(m) ) {
1240: *head = p; p = 0;
1241: } else {
1242: m0 = m;
1243: BDY(p) = NEXT(m); LEN(p)--;
1244: NEXT(m0) = 0;
1245: MKND(NV(p),m0,1,r);
1246: *head = r;
1247: }
1248: return p;
1249: }
1250:
1.1 noro 1251: PGeoBucket create_pbucket()
1252: {
1253: PGeoBucket g;
1254:
1255: g = CALLOC(1,sizeof(struct oPGeoBucket));
1256: g->m = -1;
1257: return g;
1258: }
1259:
1.25 noro 1260: void free_pbucket(PGeoBucket b) {
1261: int i;
1262:
1.26 noro 1263: for ( i = 0; i <= b->m; i++ )
1.25 noro 1264: if ( b->body[i] ) {
1265: nd_free(b->body[i]);
1266: b->body[i] = 0;
1267: }
1268: GC_free(b);
1269: }
1270:
1.63 noro 1271: void add_pbucket_symbolic(PGeoBucket g,ND d)
1272: {
1273: int l,i,k,m;
1274:
1275: if ( !d )
1276: return;
1277: l = LEN(d);
1278: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1279: /* 2^(k-1) < l <= 2^k (=m) */
1280: d = nd_merge(g->body[k],d);
1281: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1282: g->body[k] = 0;
1283: d = nd_merge(g->body[k+1],d);
1284: }
1285: g->body[k] = d;
1286: g->m = MAX(g->m,k);
1287: }
1288:
1.31 noro 1289: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1290: {
1.31 noro 1291: int l,i,k,m;
1.1 noro 1292:
1.31 noro 1293: if ( !d )
1294: return;
1295: l = LEN(d);
1.29 noro 1296: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1297: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1298: d = nd_add(mod,g->body[k],d);
1299: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1300: g->body[k] = 0;
1.31 noro 1301: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1302: }
1303: g->body[k] = d;
1304: g->m = MAX(g->m,k);
1305: }
1306:
1.26 noro 1307: void mulq_pbucket(PGeoBucket g,Q c)
1308: {
1309: int k;
1310:
1311: for ( k = 0; k <= g->m; k++ )
1312: nd_mul_c_q(g->body[k],c);
1313: }
1314:
1.63 noro 1315: NM remove_head_pbucket_symbolic(PGeoBucket g)
1316: {
1317: int j,i,k,c;
1318: NM head;
1319:
1320: k = g->m;
1321: j = -1;
1322: for ( i = 0; i <= k; i++ ) {
1323: if ( !g->body[i] ) continue;
1324: if ( j < 0 ) j = i;
1325: else {
1326: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1327: if ( c > 0 )
1328: j = i;
1329: else if ( c == 0 )
1330: g->body[i] = nd_remove_head(g->body[i]);
1331: }
1332: }
1333: if ( j < 0 ) return 0;
1334: else {
1335: head = BDY(g->body[j]);
1336: if ( !NEXT(head) ) {
1337: FREEND(g->body[j]);
1338: g->body[j] = 0;
1339: } else {
1340: BDY(g->body[j]) = NEXT(head);
1341: LEN(g->body[j])--;
1342: }
1343: return head;
1344: }
1345: }
1346:
1.19 noro 1347: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1348: {
1349: int j,i,c,k,nv,sum;
1.61 noro 1350: UINT *di,*dj;
1.1 noro 1351: ND gi,gj;
1352:
1353: k = g->m;
1354: while ( 1 ) {
1355: j = -1;
1356: for ( i = 0; i <= k; i++ ) {
1357: if ( !(gi = g->body[i]) )
1358: continue;
1359: if ( j < 0 ) {
1360: j = i;
1361: gj = g->body[j];
1362: dj = HDL(gj);
1.14 noro 1363: sum = HCM(gj);
1.1 noro 1364: } else {
1.34 noro 1365: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1366: if ( c > 0 ) {
1.34 noro 1367: if ( sum ) HCM(gj) = sum;
1368: else g->body[j] = nd_remove_head(gj);
1.1 noro 1369: j = i;
1370: gj = g->body[j];
1371: dj = HDL(gj);
1.14 noro 1372: sum = HCM(gj);
1.1 noro 1373: } else if ( c == 0 ) {
1.71 noro 1374: if ( mod == -1 )
1375: sum = _addsf(sum,HCM(gi));
1376: else {
1377: sum = sum+HCM(gi)-mod;
1378: if ( sum < 0 ) sum += mod;
1379: }
1.1 noro 1380: g->body[i] = nd_remove_head(gi);
1381: }
1382: }
1383: }
1.34 noro 1384: if ( j < 0 ) return -1;
1.1 noro 1385: else if ( sum ) {
1.14 noro 1386: HCM(gj) = sum;
1.26 noro 1387: return j;
1.31 noro 1388: } else
1.26 noro 1389: g->body[j] = nd_remove_head(gj);
1390: }
1391: }
1392:
1393: int head_pbucket_q(PGeoBucket g)
1394: {
1395: int j,i,c,k,nv;
1396: Q sum,t;
1397: ND gi,gj;
1398:
1399: k = g->m;
1400: while ( 1 ) {
1401: j = -1;
1402: for ( i = 0; i <= k; i++ ) {
1.34 noro 1403: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1404: if ( j < 0 ) {
1405: j = i;
1406: gj = g->body[j];
1407: sum = HCQ(gj);
1408: } else {
1409: nv = NV(gi);
1.34 noro 1410: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1411: if ( c > 0 ) {
1.34 noro 1412: if ( sum ) HCQ(gj) = sum;
1413: else g->body[j] = nd_remove_head(gj);
1.26 noro 1414: j = i;
1415: gj = g->body[j];
1416: sum = HCQ(gj);
1417: } else if ( c == 0 ) {
1418: addq(sum,HCQ(gi),&t);
1419: sum = t;
1420: g->body[i] = nd_remove_head(gi);
1421: }
1422: }
1423: }
1.34 noro 1424: if ( j < 0 ) return -1;
1.26 noro 1425: else if ( sum ) {
1426: HCQ(gj) = sum;
1.1 noro 1427: return j;
1.31 noro 1428: } else
1.1 noro 1429: g->body[j] = nd_remove_head(gj);
1430: }
1431: }
1432:
1.25 noro 1433: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1434: {
1.31 noro 1435: int i;
1.1 noro 1436: ND r,t;
1437:
1438: r = 0;
1.28 noro 1439: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1440: r = nd_add(mod,r,g->body[i]);
1.28 noro 1441: g->body[i] = 0;
1442: }
1443: g->m = -1;
1.1 noro 1444: return r;
1445: }
1446:
1.92 noro 1447: void do_diagonalize(int sugar,int m)
1448: {
1449: int i,nh,stat;
1450: NODE r,g,t;
1451: ND h,nf,s,head;
1452: NDV nfv;
1453: Q q,num,den;
1454: union oNDC dn;
1455:
1456: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1457: if ( nd_demand )
1458: nfv = ndv_load(i);
1459: else
1460: nfv = nd_ps[i];
1461: s = ndvtond(m,nfv);
1462: s = nd_separate_head(s,&head);
1463: nd_nf(m,s,nd_ps,1,&dn,&nf);
1464: if ( !m ) {
1465: NTOQ(NM(dn.z),SGN(dn.z),num);
1466: mulq(HCQ(head),num,&q); HCQ(head) = q;
1467: if ( DN(dn.z) ) {
1468: NTOQ(DN(dn.z),1,den);
1469: nd_mul_c_q(nf,den);
1470: }
1471: }
1472: nf = nd_add(m,head,nf);
1473: ndv_free(nfv);
1474: nd_removecont(m,nf);
1475: nfv = ndtondv(m,nf);
1476: nd_free(nf);
1477: nd_bound[i] = ndv_compute_bound(nfv);
1478: if ( nd_demand ) {
1479: ndv_save(nfv,i);
1480: ndv_free(nfv);
1481: } else
1482: nd_ps[i] = nfv;
1483: }
1484: }
1485:
1.27 noro 1486: /* return value = 0 => input is not a GB */
1487:
1.77 noro 1488: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1489: {
1490: int i,nh,sugar,stat;
1.23 noro 1491: NODE r,g,t;
1.1 noro 1492: ND_pairs d;
1493: ND_pairs l;
1.77 noro 1494: ND h,nf,s,head;
1.63 noro 1495: NDV nfv;
1.85 noro 1496: Q q,num,den;
1.69 noro 1497: union oNDC dn;
1.1 noro 1498:
1.23 noro 1499: g = 0; d = 0;
1500: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1501: d = update_pairs(d,g,i);
1502: g = update_base(g,i);
1503: }
1504: sugar = 0;
1505: while ( d ) {
1506: again:
1507: l = nd_minp(d,&d);
1.14 noro 1508: if ( SG(l) != sugar ) {
1.92 noro 1509: if ( ishomo ) do_diagonalize(sugar,m);
1510:
1.14 noro 1511: sugar = SG(l);
1.71 noro 1512: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1513: }
1.53 noro 1514: stat = nd_sp(m,0,l,&h);
1.1 noro 1515: if ( !stat ) {
1516: NEXT(l) = d; d = l;
1.20 noro 1517: d = nd_reconstruct(m,0,d);
1.1 noro 1518: goto again;
1519: }
1.41 noro 1520: #if USE_GEOBUCKET
1.69 noro 1521: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1522: #else
1.69 noro 1523: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1524: #endif
1.1 noro 1525: if ( !stat ) {
1526: NEXT(l) = d; d = l;
1.20 noro 1527: d = nd_reconstruct(m,0,d);
1.1 noro 1528: goto again;
1529: } else if ( nf ) {
1.27 noro 1530: if ( checkonly ) return 0;
1.71 noro 1531: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1532: nd_removecont(m,nf);
1533: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1534: nh = ndv_newps(m,nfv,0);
1.1 noro 1535: d = update_pairs(d,g,nh);
1536: g = update_base(g,nh);
1537: FREENDP(l);
1538: } else {
1.71 noro 1539: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1540: FREENDP(l);
1541: }
1542: }
1.77 noro 1543: if ( nd_demand )
1544: for ( t = g; t; t = NEXT(t) )
1545: BDY(t) = (pointer)ndv_load((int)BDY(t));
1546: else
1547: for ( t = g; t; t = NEXT(t) )
1548: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1549: return g;
1550: }
1551:
1.92 noro 1552: void do_diagonalize_trace(int sugar,int m)
1.91 noro 1553: {
1554: int i,nh,stat;
1555: NODE r,g,t;
1556: ND h,nf,nfq,s,head;
1557: NDV nfv,nfqv;
1558: Q q,den,num;
1559: union oNDC dn;
1560:
1561: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1562: /* for nd_ps */
1563: s = ndvtond(m,nd_ps[i]);
1564: s = nd_separate_head(s,&head);
1565: nd_nf_pbucket(m,s,nd_ps,1,&nf);
1566: nf = nd_add(m,head,nf);
1567: ndv_free(nd_ps[i]);
1568: nd_ps[i] = ndtondv(m,nf);
1569: nd_free(nf);
1570:
1571: /* for nd_ps_trace */
1572: if ( nd_demand )
1573: nfv = ndv_load(i);
1574: else
1575: nfv = nd_ps_trace[i];
1576: s = ndvtond(0,nfv);
1577: s = nd_separate_head(s,&head);
1578: nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
1579: NTOQ(NM(dn.z),SGN(dn.z),num);
1580: mulq(HCQ(head),num,&q); HCQ(head) = q;
1581: if ( DN(dn.z) ) {
1582: NTOQ(DN(dn.z),1,den);
1583: nd_mul_c_q(nf,den);
1584: }
1585: nf = nd_add(0,head,nf);
1586: ndv_free(nfv);
1587: nd_removecont(0,nf);
1588: nfv = ndtondv(0,nf);
1589: nd_free(nf);
1590: nd_bound[i] = ndv_compute_bound(nfv);
1591: if ( nd_demand ) {
1592: ndv_save(nfv,i);
1593: ndv_free(nfv);
1594: } else
1595: nd_ps_trace[i] = nfv;
1596: }
1597: }
1598:
1.82 noro 1599: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1600: {
1601: int i,nh,sugar,stat;
1.23 noro 1602: NODE r,g,t;
1.20 noro 1603: ND_pairs d;
1604: ND_pairs l;
1.82 noro 1605: ND h,nf,nfq,s,head;
1.63 noro 1606: NDV nfv,nfqv;
1.85 noro 1607: Q q,den,num;
1.69 noro 1608: union oNDC dn;
1.20 noro 1609:
1.23 noro 1610: g = 0; d = 0;
1611: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1612: d = update_pairs(d,g,i);
1613: g = update_base(g,i);
1614: }
1615: sugar = 0;
1616: while ( d ) {
1617: again:
1618: l = nd_minp(d,&d);
1619: if ( SG(l) != sugar ) {
1.93 noro 1620: if ( ishomo ) do_diagonalize_trace(sugar,m);
1.20 noro 1621: sugar = SG(l);
1.71 noro 1622: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1623: }
1.53 noro 1624: stat = nd_sp(m,0,l,&h);
1.20 noro 1625: if ( !stat ) {
1626: NEXT(l) = d; d = l;
1627: d = nd_reconstruct(m,1,d);
1628: goto again;
1629: }
1.41 noro 1630: #if USE_GEOBUCKET
1.53 noro 1631: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1632: #else
1.69 noro 1633: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1634: #endif
1.20 noro 1635: if ( !stat ) {
1636: NEXT(l) = d; d = l;
1637: d = nd_reconstruct(m,1,d);
1638: goto again;
1639: } else if ( nf ) {
1.77 noro 1640: if ( nd_demand ) {
1641: nfqv = ndv_load(nd_psn);
1642: nfq = ndvtond(0,nfqv);
1643: } else
1644: nfq = 0;
1645: if ( !nfq ) {
1646: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1647: NEXT(l) = d; d = l;
1648: d = nd_reconstruct(m,1,d);
1649: goto again;
1650: }
1.70 noro 1651: }
1.20 noro 1652: if ( nfq ) {
1.70 noro 1653: /* m|HC(nfq) => failure */
1.63 noro 1654: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1655:
1.71 noro 1656: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1657: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1658: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 noro 1659: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1660: d = update_pairs(d,g,nh);
1661: g = update_base(g,nh);
1662: } else {
1.71 noro 1663: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1664: }
1665: } else {
1.71 noro 1666: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1667: }
1668: FREENDP(l);
1669: }
1.77 noro 1670: if ( nd_demand )
1671: for ( t = g; t; t = NEXT(t) )
1672: BDY(t) = (pointer)ndv_load((int)BDY(t));
1673: else
1674: for ( t = g; t; t = NEXT(t) )
1675: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1676: return g;
1677: }
1678:
1.23 noro 1679: int ndv_compare(NDV *p1,NDV *p2)
1680: {
1.34 noro 1681: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1682: }
1683:
1684: int ndv_compare_rev(NDV *p1,NDV *p2)
1685: {
1.34 noro 1686: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1687: }
1688:
1.61 noro 1689: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1690: {
1.69 noro 1691: int i,n,stat;
1692: ND nf,g,head;
1.23 noro 1693: NODE t,a0,a;
1.69 noro 1694: union oNDC dn;
1695: NDV *w;
1.85 noro 1696: Q q,num,den;
1.23 noro 1697:
1.69 noro 1698: n = length(f);
1699: #if 0
1700: w = (NDV *)ALLOCA(n*sizeof(NDV));
1701: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1702: qsort(w,n,sizeof(NDV),
1703: (int (*)(const void *,const void *))ndv_compare);
1704: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1705: #endif
1706: ndv_setup(m,0,f);
1707: for ( i = 0; i < n; ) {
1708: g = ndvtond(m,nd_ps[i]);
1709: g = nd_separate_head(g,&head);
1710: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1711: if ( !stat )
1.69 noro 1712: nd_reconstruct(m,0,0);
1713: else {
1.71 noro 1714: if ( DP_Print ) { printf("."); fflush(stdout); }
1.85 noro 1715: if ( !m ) {
1716: NTOQ(NM(dn.z),SGN(dn.z),num);
1717: mulq(HCQ(head),num,&q); HCQ(head) = q;
1718: if ( DN(dn.z) ) {
1719: NTOQ(DN(dn.z),1,den);
1720: nd_mul_c_q(nf,den);
1721: }
1722: }
1.69 noro 1723: nf = nd_add(m,head,nf);
1724: ndv_free(nd_ps[i]);
1.24 noro 1725: nd_removecont(m,nf);
1.69 noro 1726: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1727: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1728: i++;
1.23 noro 1729: }
1730: }
1.71 noro 1731: if ( DP_Print ) { printf("\n"); }
1.23 noro 1732: for ( a0 = 0, i = 0; i < n; i++ ) {
1733: NEXTNODE(a0,a);
1.69 noro 1734: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1735: }
1736: NEXT(a) = 0;
1737: return a0;
1738: }
1739:
1.1 noro 1740: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1741: {
1742: ND_pairs d1,nd,cur,head,prev,remove;
1743:
1744: if ( !g ) return d;
1745: d = crit_B(d,t);
1746: d1 = nd_newpairs(g,t);
1747: d1 = crit_M(d1);
1748: d1 = crit_F(d1);
1.55 noro 1749: if ( do_weyl )
1750: head = d1;
1751: else {
1752: prev = 0; cur = head = d1;
1753: while ( cur ) {
1754: if ( crit_2( cur->i1,cur->i2 ) ) {
1755: remove = cur;
1756: if ( !prev ) head = cur = NEXT(cur);
1757: else cur = NEXT(prev) = NEXT(cur);
1758: FREENDP(remove);
1759: } else {
1760: prev = cur; cur = NEXT(cur);
1761: }
1.1 noro 1762: }
1763: }
1764: if ( !d )
1765: return head;
1766: else {
1767: nd = d;
1.34 noro 1768: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1769: NEXT(nd) = head;
1770: return d;
1771: }
1772: }
1773:
1774: ND_pairs nd_newpairs( NODE g, int t )
1775: {
1776: NODE h;
1.61 noro 1777: UINT *dl;
1.34 noro 1778: int ts,s;
1.1 noro 1779: ND_pairs r,r0;
1780:
1.20 noro 1781: dl = DL(nd_psh[t]);
1.34 noro 1782: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1783: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1784: NEXTND_pairs(r0,r);
1785: r->i1 = (int)BDY(h);
1786: r->i2 = t;
1.20 noro 1787: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1788: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1789: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1790: }
1791: NEXT(r) = 0;
1792: return r0;
1793: }
1794:
1795: ND_pairs crit_B( ND_pairs d, int s )
1796: {
1797: ND_pairs cur,head,prev,remove;
1.61 noro 1798: UINT *t,*tl,*lcm;
1.1 noro 1799: int td,tdl;
1800:
1801: if ( !d ) return 0;
1.20 noro 1802: t = DL(nd_psh[s]);
1.1 noro 1803: prev = 0;
1804: head = cur = d;
1.61 noro 1805: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1806: while ( cur ) {
1807: tl = cur->lcm;
1808: if ( ndl_reducible(tl,t)
1.20 noro 1809: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1810: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1811: remove = cur;
1812: if ( !prev ) {
1813: head = cur = NEXT(cur);
1814: } else {
1815: cur = NEXT(prev) = NEXT(cur);
1816: }
1817: FREENDP(remove);
1818: } else {
1.34 noro 1819: prev = cur; cur = NEXT(cur);
1.1 noro 1820: }
1821: }
1822: return head;
1823: }
1824:
1825: ND_pairs crit_M( ND_pairs d1 )
1826: {
1827: ND_pairs e,d2,d3,dd,p;
1.61 noro 1828: UINT *id,*jd;
1.1 noro 1829:
1830: for ( dd = 0, e = d1; e; e = d3 ) {
1831: if ( !(d2 = NEXT(e)) ) {
1832: NEXT(e) = dd;
1833: return e;
1834: }
1.34 noro 1835: id = LCM(e);
1.1 noro 1836: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 1837: p = NEXT(d2);
1838: jd = LCM(d2);
1839: if ( ndl_equal(jd,id) )
1840: ;
1841: else if ( TD(jd) > TD(id) )
1.1 noro 1842: if ( ndl_reducible(jd,id) ) continue;
1843: else ;
1.34 noro 1844: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 1845: NEXT(d2) = d3;
1846: d3 = d2;
1847: }
1848: NEXT(e) = dd;
1849: dd = e;
1850: continue;
1851: /**/
1852: delit: NEXT(d2) = d3;
1853: d3 = d2;
1854: for ( ; p; p = d2 ) {
1855: d2 = NEXT(p);
1856: NEXT(p) = d3;
1857: d3 = p;
1858: }
1859: FREENDP(e);
1860: }
1861: return dd;
1862: }
1863:
1864: ND_pairs crit_F( ND_pairs d1 )
1865: {
1866: ND_pairs rest, head,remove;
1867: ND_pairs last, p, r, w;
1868: int s;
1869:
1870: for ( head = last = 0, p = d1; NEXT(p); ) {
1871: r = w = equivalent_pairs(p,&rest);
1.14 noro 1872: s = SG(r);
1.1 noro 1873: w = NEXT(w);
1874: while ( w ) {
1875: if ( crit_2(w->i1,w->i2) ) {
1876: r = w;
1877: w = NEXT(w);
1878: while ( w ) {
1879: remove = w;
1880: w = NEXT(w);
1881: FREENDP(remove);
1882: }
1883: break;
1.14 noro 1884: } else if ( SG(w) < s ) {
1.1 noro 1885: FREENDP(r);
1886: r = w;
1.14 noro 1887: s = SG(r);
1.1 noro 1888: w = NEXT(w);
1889: } else {
1890: remove = w;
1891: w = NEXT(w);
1892: FREENDP(remove);
1893: }
1894: }
1895: if ( last ) NEXT(last) = r;
1896: else head = r;
1897: NEXT(last = r) = 0;
1898: p = rest;
1899: if ( !p ) return head;
1900: }
1901: if ( !last ) return p;
1902: NEXT(last) = p;
1903: return head;
1904: }
1905:
1906: int crit_2( int dp1, int dp2 )
1907: {
1.20 noro 1908: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 1909: }
1910:
1.40 noro 1911: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 1912: {
1913: ND_pairs w,p,r,s;
1.61 noro 1914: UINT *d;
1.1 noro 1915:
1916: w = d1;
1.34 noro 1917: d = LCM(w);
1.1 noro 1918: s = NEXT(w);
1919: NEXT(w) = 0;
1920: for ( r = 0; s; s = p ) {
1921: p = NEXT(s);
1.34 noro 1922: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 1923: NEXT(s) = w; w = s;
1.1 noro 1924: } else {
1.39 noro 1925: NEXT(s) = r; r = s;
1.1 noro 1926: }
1927: }
1928: *prest = r;
1929: return w;
1930: }
1931:
1932: NODE update_base(NODE nd,int ndp)
1933: {
1.61 noro 1934: UINT *dl, *dln;
1.1 noro 1935: NODE last, p, head;
1936:
1.20 noro 1937: dl = DL(nd_psh[ndp]);
1.1 noro 1938: for ( head = last = 0, p = nd; p; ) {
1.20 noro 1939: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 1940: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 1941: p = NEXT(p);
1942: if ( last ) NEXT(last) = p;
1943: } else {
1944: if ( !last ) head = p;
1945: p = NEXT(last = p);
1946: }
1947: }
1948: head = append_one(head,ndp);
1949: return head;
1950: }
1951:
1952: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
1953: {
1954: ND_pairs m,ml,p,l;
1.61 noro 1955: UINT *lcm;
1.33 noro 1956: int s,td,len,tlen,c,c1;
1.1 noro 1957:
1958: if ( !(p = NEXT(m = d)) ) {
1959: *prest = p;
1960: NEXT(m) = 0;
1961: return m;
1962: }
1.14 noro 1963: s = SG(m);
1.77 noro 1964: if ( !NoSugar ) {
1965: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1966: if ( (SG(p) < s)
1967: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
1968: ml = l; m = p; s = SG(m);
1969: }
1970: } else {
1971: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1972: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
1973: ml = l; m = p; s = SG(m);
1974: }
1975: }
1.1 noro 1976: if ( !ml ) *prest = NEXT(m);
1977: else {
1978: NEXT(ml) = NEXT(m);
1979: *prest = d;
1980: }
1981: NEXT(m) = 0;
1982: return m;
1983: }
1984:
1.63 noro 1985: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
1986: {
1.69 noro 1987: int msugar,i;
1.63 noro 1988: ND_pairs t,dm0,dm,dr0,dr;
1989:
1990: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
1991: if ( SG(t) < msugar ) msugar = SG(t);
1992: dm0 = 0; dr0 = 0;
1.69 noro 1993: for ( i = 0, t = d; t; t = NEXT(t) )
1994: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 1995: if ( dm0 ) NEXT(dm) = t;
1996: else dm0 = t;
1997: dm = t;
1.69 noro 1998: i++;
1.63 noro 1999: } else {
2000: if ( dr0 ) NEXT(dr) = t;
2001: else dr0 = t;
2002: dr = t;
2003: }
2004: NEXT(dm) = 0;
2005: if ( dr0 ) NEXT(dr) = 0;
2006: *prest = dr0;
2007: return dm0;
2008: }
2009:
1.77 noro 2010: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2011: {
1.3 noro 2012: int len;
1.13 noro 2013: RHist r;
1.20 noro 2014: NDV b;
1.3 noro 2015:
1.1 noro 2016: if ( nd_psn == nd_pslen ) {
2017: nd_pslen *= 2;
1.11 noro 2018: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2019: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2020: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2021: nd_bound = (UINT **)
2022: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2023: }
1.39 noro 2024: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2025: nd_ps[nd_psn] = a;
1.39 noro 2026: if ( aq ) {
1.63 noro 2027: nd_ps_trace[nd_psn] = aq;
2028: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2029: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2030: } else {
1.63 noro 2031: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2032: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2033: }
1.77 noro 2034: if ( nd_demand ) {
2035: if ( aq ) {
2036: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2037: nd_ps_trace[nd_psn] = 0;
2038: } else {
2039: ndv_save(nd_ps[nd_psn],nd_psn);
2040: nd_ps[nd_psn] = 0;
2041: }
2042: }
1.1 noro 2043: return nd_psn++;
2044: }
2045:
1.61 noro 2046: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2047: {
1.5 noro 2048: int i,j,td,len,max;
1.1 noro 2049: NODE s,s0,f0;
1.61 noro 2050: UINT *d;
1.13 noro 2051: RHist r;
1.69 noro 2052: NDV *w;
1.61 noro 2053: NDV a,am;
1.11 noro 2054:
2055: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2056:
1.69 noro 2057: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2058: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2059: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2060: qsort(w,nd_psn,sizeof(NDV),
2061: (int (*)(const void *,const void *))ndv_compare);
2062: nd_pslen = 2*nd_psn;
1.11 noro 2063: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2064: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2065: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2066: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2067:
1.11 noro 2068: if ( !nd_red )
1.13 noro 2069: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2070: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2071: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2072: if ( trace ) {
1.69 noro 2073: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2074: ndv_removecont(0,a);
2075: am = nd_ps[i] = ndv_dup(mod,a);
2076: ndv_mod(mod,am);
2077: ndv_removecont(mod,am);
2078: } else {
1.69 noro 2079: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2080: ndv_removecont(mod,a);
1.39 noro 2081: }
1.61 noro 2082: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2083: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2084: nd_psh[i] = r;
1.77 noro 2085: if ( nd_demand ) {
2086: if ( trace ) {
2087: ndv_save(nd_ps_trace[i],i);
2088: nd_ps_trace[i] = 0;
2089: } else {
2090: ndv_save(nd_ps[i],i);
2091: nd_ps[i] = 0;
2092: }
2093: }
1.20 noro 2094: }
2095: }
2096:
1.63 noro 2097: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2098: {
1.61 noro 2099: VL tv,fv,vv,vc;
1.1 noro 2100: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2101: int e,max,nvar;
1.69 noro 2102: NDV b;
1.77 noro 2103: int ishomo;
1.1 noro 2104:
1.78 noro 2105: if ( !m && Demand ) nd_demand = 1;
2106: else nd_demand = 0;
2107:
1.74 noro 2108: ndv_alloc = 0;
1.1 noro 2109: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2110: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2111: switch ( ord->id ) {
2112: case 1:
2113: if ( ord->nv != nvar )
2114: error("nd_{gr,f4} : invalid order specification");
2115: break;
2116: default:
2117: break;
2118: }
1.32 noro 2119: nd_init_ord(ord);
1.61 noro 2120: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2121: for ( tv = vv; tv; tv = NEXT(tv) ) {
2122: e = getdeg(tv->v,(P)BDY(t));
2123: max = MAX(e,max);
2124: }
2125: nd_setup_parameters(nvar,max);
1.77 noro 2126: ishomo = 1;
1.1 noro 2127: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2128: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2129: if ( ishomo )
2130: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2131: if ( m ) ndv_mod(m,b);
1.61 noro 2132: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2133: }
2134: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2135: ndv_setup(m,0,fd0);
1.77 noro 2136: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 noro 2137: nd_demand = 0;
1.61 noro 2138: x = ndv_reducebase(x);
2139: x = ndv_reduceall(m,x);
1.23 noro 2140: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2141: NEXTNODE(r0,r);
1.61 noro 2142: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2143: }
2144: if ( r0 ) NEXT(r) = 0;
2145: MKLIST(*rp,r0);
1.74 noro 2146: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2147: }
2148:
1.52 noro 2149: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2150: {
1.89 noro 2151: struct order_spec *ord1;
1.61 noro 2152: VL tv,fv,vv,vc;
1.27 noro 2153: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2154: int m,nocheck,nvar,mindex,e,max;
2155: NDV c;
2156: NMV a;
1.27 noro 2157: P p;
1.61 noro 2158: EPOS oepos;
1.82 noro 2159: int obpe,oadv,wmax,i,len,cbpe,ishomo;
1.20 noro 2160:
2161: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2162: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.88 noro 2163: switch ( ord->id ) {
2164: case 1:
2165: if ( ord->nv != nvar )
2166: error("nd_gr_trace : invalid order specification");
2167: break;
2168: default:
2169: break;
2170: }
1.52 noro 2171: nocheck = 0;
2172: mindex = 0;
2173:
1.78 noro 2174: if ( Demand ) nd_demand = 1;
2175: else nd_demand = 0;
2176:
1.52 noro 2177: /* setup modulus */
2178: if ( trace < 0 ) {
2179: trace = -trace;
2180: nocheck = 1;
2181: }
2182: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2183: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2184: for ( tv = vv; tv; tv = NEXT(tv) ) {
2185: e = getdeg(tv->v,(P)BDY(t));
2186: max = MAX(e,max);
1.23 noro 2187: }
1.61 noro 2188: nd_init_ord(ord);
2189: nd_setup_parameters(nvar,max);
2190: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 noro 2191: ishomo = 1;
1.61 noro 2192: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2193: c = ptondv(CO,vv,(P)BDY(t));
1.82 noro 2194: if ( ishomo )
2195: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2196: if ( c ) {
2197: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2198: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2199: }
1.61 noro 2200: }
2201: if ( in0 ) NEXT(in) = 0;
2202: if ( fd0 ) NEXT(fd) = 0;
1.82 noro 2203: if ( !ishomo && homo ) {
1.61 noro 2204: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2205: c = (NDV)BDY(t); len = LEN(c);
2206: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2207: wmax = MAX(TD(DL(a)),wmax);
2208: }
2209: homogenize_order(ord,nvar,&ord1);
1.89 noro 2210: nd_init_ord(ord1);
1.61 noro 2211: nd_setup_parameters(nvar+1,wmax);
2212: for ( t = fd0; t; t = NEXT(t) )
2213: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2214: }
1.52 noro 2215: while ( 1 ) {
1.77 noro 2216: if ( Demand )
2217: nd_demand = 1;
1.61 noro 2218: ndv_setup(m,1,fd0);
1.82 noro 2219: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2220: if ( !cand ) {
2221: /* failure */
1.61 noro 2222: if ( trace > 1 ) { *rp = 0; return; }
2223: else m = get_lprime(++mindex);
1.52 noro 2224: continue;
2225: }
1.82 noro 2226: if ( !ishomo && homo ) {
1.27 noro 2227: /* dehomogenization */
1.61 noro 2228: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2229: nd_init_ord(ord);
1.61 noro 2230: nd_setup_parameters(nvar,0);
1.27 noro 2231: }
1.77 noro 2232: nd_demand = 0;
1.61 noro 2233: cand = ndv_reducebase(cand);
2234: cand = ndv_reduceall(0,cand);
1.78 noro 2235: cbpe = nd_bpe;
1.61 noro 2236: if ( nocheck )
2237: break;
2238: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2239: /* success */
2240: break;
2241: else if ( trace > 1 ) {
2242: /* failure */
2243: *rp = 0; return;
1.61 noro 2244: } else {
1.52 noro 2245: /* try the next modulus */
2246: m = get_lprime(++mindex);
1.61 noro 2247: /* reset the parameters */
1.82 noro 2248: if ( !ishomo && homo ) {
1.89 noro 2249: nd_init_ord(ord1);
1.61 noro 2250: nd_setup_parameters(nvar+1,wmax);
2251: } else {
2252: nd_init_ord(ord);
2253: nd_setup_parameters(nvar,max);
2254: }
2255: }
1.52 noro 2256: }
1.27 noro 2257: /* dp->p */
1.61 noro 2258: nd_bpe = cbpe;
1.73 noro 2259: nd_setup_parameters(nd_nvar,0);
1.61 noro 2260: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2261: MKLIST(*rp,cand);
1.1 noro 2262: }
2263:
1.61 noro 2264: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2265: {
1.61 noro 2266: UINT *d;
1.57 noro 2267: int i,j,l,s,ord_l;
1.43 noro 2268: struct order_pair *op;
1.1 noro 2269:
2270: d = dl->d;
1.41 noro 2271: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2272: if ( nd_blockmask ) {
2273: l = nd_blockmask->n;
2274: op = nd_blockmask->order_pair;
2275: for ( j = 0, s = 0; j < l; j++ ) {
2276: ord_l = op[j].length;
1.57 noro 2277: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2278: }
2279: TD(r) = ndl_weight(r);
1.61 noro 2280: ndl_weight_mask(r);
1.43 noro 2281: } else {
1.56 noro 2282: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2283: TD(r) = ndl_weight(r);
2284: }
1.1 noro 2285: }
2286:
1.61 noro 2287: DL ndltodl(int n,UINT *ndl)
1.1 noro 2288: {
2289: DL dl;
2290: int *d;
1.57 noro 2291: int i,j,l,s,ord_l;
1.43 noro 2292: struct order_pair *op;
1.1 noro 2293:
2294: NEWDL(dl,n);
1.34 noro 2295: dl->td = TD(ndl);
1.1 noro 2296: d = dl->d;
1.43 noro 2297: if ( nd_blockmask ) {
2298: l = nd_blockmask->n;
2299: op = nd_blockmask->order_pair;
2300: for ( j = 0, s = 0; j < l; j++ ) {
2301: ord_l = op[j].length;
1.57 noro 2302: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2303: }
2304: } else {
1.56 noro 2305: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2306: }
1.1 noro 2307: return dl;
2308: }
2309:
1.61 noro 2310: void ndl_print(UINT *dl)
1.1 noro 2311: {
2312: int n;
1.57 noro 2313: int i,j,l,ord_l,s,s0;
1.43 noro 2314: struct order_pair *op;
1.1 noro 2315:
2316: n = nd_nvar;
2317: printf("<<");
1.43 noro 2318: if ( nd_blockmask ) {
2319: l = nd_blockmask->n;
2320: op = nd_blockmask->order_pair;
2321: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2322: ord_l = op[j].length;
1.57 noro 2323: for ( i = 0; i < ord_l; i++, s++ )
2324: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2325: }
2326: } else {
1.56 noro 2327: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2328: }
1.1 noro 2329: printf(">>");
2330: }
2331:
2332: void nd_print(ND p)
2333: {
2334: NM m;
2335:
2336: if ( !p )
2337: printf("0\n");
2338: else {
2339: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2340: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2341: else printf("+%d*",CM(m));
1.14 noro 2342: ndl_print(DL(m));
1.1 noro 2343: }
2344: printf("\n");
2345: }
2346: }
2347:
1.16 noro 2348: void nd_print_q(ND p)
2349: {
2350: NM m;
2351:
2352: if ( !p )
2353: printf("0\n");
2354: else {
2355: for ( m = BDY(p); m; m = NEXT(m) ) {
2356: printf("+");
1.74 noro 2357: printexpr(CO,(Obj)CQ(m));
1.16 noro 2358: printf("*");
2359: ndl_print(DL(m));
2360: }
2361: printf("\n");
2362: }
2363: }
2364:
1.1 noro 2365: void ndp_print(ND_pairs d)
2366: {
2367: ND_pairs t;
2368:
1.34 noro 2369: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2370: printf("\n");
2371: }
2372:
1.20 noro 2373: void nd_removecont(int mod,ND p)
1.16 noro 2374: {
2375: int i,n;
2376: Q *w;
2377: Q dvr,t;
2378: NM m;
1.21 noro 2379: struct oVECT v;
2380: N q,r;
1.16 noro 2381:
1.71 noro 2382: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2383: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2384: else {
2385: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2386: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2387: v.len = n;
2388: v.body = (pointer *)w;
1.34 noro 2389: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2390: removecont_array(w,n);
2391: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2392: }
2393: }
2394:
1.21 noro 2395: void nd_removecont2(ND p1,ND p2)
2396: {
2397: int i,n1,n2,n;
2398: Q *w;
2399: Q dvr,t;
2400: NM m;
2401: struct oVECT v;
2402: N q,r;
2403:
2404: if ( !p1 ) {
2405: nd_removecont(0,p2); return;
2406: } else if ( !p2 ) {
2407: nd_removecont(0,p1); return;
2408: }
2409: n1 = nd_length(p1);
2410: n2 = nd_length(p2);
2411: n = n1+n2;
2412: w = (Q *)ALLOCA(n*sizeof(Q));
2413: v.len = n;
2414: v.body = (pointer *)w;
1.34 noro 2415: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2416: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2417: removecont_array(w,n);
2418: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2419: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2420: }
2421:
1.20 noro 2422: void ndv_removecont(int mod,NDV p)
1.16 noro 2423: {
2424: int i,len;
2425: Q *w;
2426: Q dvr,t;
2427: NMV m;
2428:
1.71 noro 2429: if ( mod == -1 )
2430: ndv_mul_c(mod,p,_invsf(HCM(p)));
2431: else if ( mod )
1.20 noro 2432: ndv_mul_c(mod,p,invm(HCM(p),mod));
2433: else {
2434: len = p->len;
2435: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2436: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2437: sortbynm(w,len);
2438: qltozl(w,len,&dvr);
2439: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2440: divq(CQ(m),dvr,&t); CQ(m) = t;
2441: }
1.16 noro 2442: }
1.21 noro 2443: }
2444:
1.61 noro 2445: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2446: {
2447: int len,i,max;
2448: NMV m,mr0,mr,t;
2449:
2450: len = p->len;
2451: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2452: max = MAX(max,TD(DL(m)));
2453: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2454: m = (NMV)((char *)mr0+(len-1)*oadv);
2455: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2456: t = (NMV)ALLOCA(nmv_adv);
2457: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2458: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2459: CQ(mr) = CQ(m);
2460: ndl_copy(DL(t),DL(mr));
2461: }
2462: NV(p)++;
2463: BDY(p) = mr0;
2464: }
2465:
1.45 noro 2466: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2467: {
1.45 noro 2468: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2469: Q *w;
2470: Q dvr,t;
2471: NMV m,r;
2472:
2473: len = p->len;
2474: newnvar = nd_nvar-1;
1.48 noro 2475: newexporigin = nd_get_exporigin(ord);
1.45 noro 2476: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2477: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2478: ndl_dehomogenize(DL(m));
1.23 noro 2479: if ( newwpd != nd_wpd ) {
1.90 noro 2480: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
1.61 noro 2481: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2482: CQ(r) = CQ(m);
2483: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2484: adj = nd_exporigin-newexporigin;
2485: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2486: }
2487: }
2488: NV(p)--;
2489: }
2490:
1.21 noro 2491: void removecont_array(Q *c,int n)
2492: {
2493: struct oVECT v;
2494: Q d0,d1,a,u,u1,gcd;
2495: int i;
2496: N qn,rn,gn;
2497: Q *q,*r;
2498:
2499: q = (Q *)ALLOCA(n*sizeof(Q));
2500: r = (Q *)ALLOCA(n*sizeof(Q));
2501: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2502: igcdv_estimate(&v,&d0);
2503: for ( i = 0; i < n; i++ ) {
2504: divn(NM(c[i]),NM(d0),&qn,&rn);
2505: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2506: NTOQ(rn,SGN(c[i]),r[i]);
2507: }
1.34 noro 2508: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2509: if ( i < n ) {
2510: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2511: igcdv(&v,&d1);
2512: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2513: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2514: for ( i = 0; i < n; i++ ) {
2515: mulq(a,q[i],&u);
2516: if ( r[i] ) {
2517: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2518: addq(u,u1,&q[i]);
2519: } else
2520: q[i] = u;
2521: }
2522: }
1.34 noro 2523: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2524: }
2525:
1.19 noro 2526: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2527: {
2528: NM m;
2529: int c,c1;
2530:
1.34 noro 2531: if ( !p ) return;
1.71 noro 2532: if ( mod == -1 )
2533: for ( m = BDY(p); m; m = NEXT(m) )
2534: CM(m) = _mulsf(CM(m),mul);
2535: else
2536: for ( m = BDY(p); m; m = NEXT(m) ) {
2537: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2538: }
1.1 noro 2539: }
2540:
1.16 noro 2541: void nd_mul_c_q(ND p,Q mul)
2542: {
2543: NM m;
2544: Q c;
2545:
1.34 noro 2546: if ( !p ) return;
1.16 noro 2547: for ( m = BDY(p); m; m = NEXT(m) ) {
2548: mulq(CQ(m),mul,&c); CQ(m) = c;
2549: }
2550: }
2551:
1.61 noro 2552: void nd_mul_c_p(VL vl,ND p,P mul)
2553: {
2554: NM m;
2555: P c;
2556:
2557: if ( !p ) return;
2558: for ( m = BDY(p); m; m = NEXT(m) ) {
2559: mulp(vl,CP(m),mul,&c); CP(m) = c;
2560: }
2561: }
2562:
1.1 noro 2563: void nd_free(ND p)
2564: {
2565: NM t,s;
2566:
1.34 noro 2567: if ( !p ) return;
1.1 noro 2568: t = BDY(p);
2569: while ( t ) {
2570: s = NEXT(t);
2571: FREENM(t);
2572: t = s;
2573: }
2574: FREEND(p);
2575: }
2576:
1.23 noro 2577: void ndv_free(NDV p)
2578: {
2579: GC_free(BDY(p));
2580: }
2581:
1.61 noro 2582: void nd_append_red(UINT *d,int i)
1.1 noro 2583: {
1.13 noro 2584: RHist m,m0;
1.1 noro 2585: int h;
2586:
1.13 noro 2587: NEWRHist(m);
1.34 noro 2588: h = ndl_hash_value(d);
1.13 noro 2589: m->index = i;
1.14 noro 2590: ndl_copy(d,DL(m));
1.1 noro 2591: NEXT(m) = nd_red[h];
2592: nd_red[h] = m;
2593: }
2594:
1.61 noro 2595: UINT *ndv_compute_bound(NDV p)
1.1 noro 2596: {
1.61 noro 2597: UINT *d1,*d2,*t;
2598: UINT u;
1.57 noro 2599: int i,j,k,l,len,ind;
1.45 noro 2600: NMV m;
1.1 noro 2601:
2602: if ( !p )
2603: return 0;
1.61 noro 2604: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2605: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2606: len = LEN(p);
2607: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2608: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2609: ndl_lcm(DL(m),d1,d2);
1.1 noro 2610: t = d1; d1 = d2; d2 = t;
2611: }
1.12 noro 2612: l = nd_nvar+31;
1.61 noro 2613: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2614: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2615: u = d1[i];
2616: k = (nd_epw-1)*nd_bpe;
2617: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2618: t[ind] = (u>>k)&nd_mask0;
2619: }
2620: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2621: return t;
2622: }
2623:
1.48 noro 2624: int nd_get_exporigin(struct order_spec *ord)
2625: {
1.51 noro 2626: switch ( ord->id ) {
1.41 noro 2627: case 0:
1.48 noro 2628: return 1;
1.41 noro 2629: case 1:
2630: /* block order */
1.43 noro 2631: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2632: return ord->ord.block.length+1;
1.41 noro 2633: case 2:
1.52 noro 2634: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2635: }
1.48 noro 2636: }
2637:
1.61 noro 2638: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2639: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2640: struct order_pair *op;
1.48 noro 2641:
1.73 noro 2642: nd_nvar = nvar;
1.79 noro 2643: if ( max ) {
1.83 noro 2644: /* XXX */
2645: if ( do_weyl ) nd_bpe = 32;
2646: else if ( max < 2 ) nd_bpe = 1;
1.79 noro 2647: else if ( max < 4 ) nd_bpe = 2;
2648: else if ( max < 8 ) nd_bpe = 3;
2649: else if ( max < 16 ) nd_bpe = 4;
2650: else if ( max < 32 ) nd_bpe = 5;
2651: else if ( max < 64 ) nd_bpe = 6;
2652: else if ( max < 256 ) nd_bpe = 8;
2653: else if ( max < 1024 ) nd_bpe = 10;
2654: else if ( max < 65536 ) nd_bpe = 16;
2655: else nd_bpe = 32;
2656: }
1.61 noro 2657: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2658: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2659: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2660: wpd = nd_exporigin+elen;
2661: if ( wpd != nd_wpd ) {
2662: nd_free_private_storage();
2663: nd_wpd = wpd;
2664: }
1.1 noro 2665: if ( nd_bpe < 32 ) {
2666: nd_mask0 = (1<<nd_bpe)-1;
2667: } else {
2668: nd_mask0 = 0xffffffff;
2669: }
2670: bzero(nd_mask,sizeof(nd_mask));
2671: nd_mask1 = 0;
2672: for ( i = 0; i < nd_epw; i++ ) {
2673: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2674: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2675: }
1.90 noro 2676: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
1.57 noro 2677: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2678: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2679: }
2680:
1.20 noro 2681: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2682: {
1.37 noro 2683: int i,obpe,oadv,h;
1.71 noro 2684: static NM prev_nm_free_list;
2685: static ND_pairs prev_ndp_free_list;
1.13 noro 2686: RHist mr0,mr;
2687: RHist r;
1.37 noro 2688: RHist *old_red;
1.71 noro 2689: ND_pairs s0,s,t;
1.43 noro 2690: EPOS oepos;
1.15 noro 2691:
1.1 noro 2692: obpe = nd_bpe;
1.11 noro 2693: oadv = nmv_adv;
1.43 noro 2694: oepos = nd_epos;
1.63 noro 2695: if ( obpe < 2 ) nd_bpe = 2;
2696: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2697: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2698: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2699: else if ( obpe < 6 ) nd_bpe = 6;
2700: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2701: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2702: else if ( obpe < 16 ) nd_bpe = 16;
2703: else if ( obpe < 32 ) nd_bpe = 32;
2704: else error("nd_reconstruct : exponent too large");
1.5 noro 2705:
1.73 noro 2706: nd_setup_parameters(nd_nvar,0);
1.1 noro 2707: prev_nm_free_list = _nm_free_list;
2708: prev_ndp_free_list = _ndp_free_list;
2709: _nm_free_list = 0;
2710: _ndp_free_list = 0;
1.53 noro 2711: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2712: if ( trace )
2713: for ( i = nd_psn-1; i >= 0; i-- )
2714: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2715: s0 = 0;
2716: for ( t = d; t; t = NEXT(t) ) {
2717: NEXTND_pairs(s0,s);
2718: s->i1 = t->i1;
2719: s->i2 = t->i2;
1.14 noro 2720: SG(s) = SG(t);
1.61 noro 2721: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2722: }
1.37 noro 2723:
2724: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2725: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2726: old_red[i] = nd_red[i];
2727: nd_red[i] = 0;
2728: }
2729: for ( i = 0; i < REDTAB_LEN; i++ )
2730: for ( r = old_red[i]; r; r = NEXT(r) ) {
2731: NEWRHist(mr);
1.13 noro 2732: mr->index = r->index;
1.20 noro 2733: SG(mr) = SG(r);
1.61 noro 2734: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2735: h = ndl_hash_value(DL(mr));
2736: NEXT(mr) = nd_red[h];
2737: nd_red[h] = mr;
1.6 noro 2738: }
1.37 noro 2739: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2740: old_red = 0;
1.11 noro 2741: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2742: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2743: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2744: nd_psh[i] = r;
1.11 noro 2745: }
1.1 noro 2746: if ( s0 ) NEXT(s) = 0;
2747: prev_nm_free_list = 0;
2748: prev_ndp_free_list = 0;
1.71 noro 2749: #if 0
1.1 noro 2750: GC_gcollect();
1.71 noro 2751: #endif
1.1 noro 2752: return s0;
2753: }
2754:
1.61 noro 2755: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2756: {
1.57 noro 2757: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2758: struct order_pair *op;
1.1 noro 2759:
2760: n = nd_nvar;
1.61 noro 2761: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2762: omask0 = (1<<obpe)-1;
1.34 noro 2763: TD(r) = TD(d);
1.41 noro 2764: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2765: if ( nd_blockmask ) {
2766: l = nd_blockmask->n;
2767: op = nd_blockmask->order_pair;
2768: for ( i = 1; i < nd_exporigin; i++ )
2769: r[i] = d[i];
2770: for ( j = 0, s = 0; j < l; j++ ) {
2771: ord_l = op[j].length;
1.57 noro 2772: for ( i = 0; i < ord_l; i++, s++ ) {
2773: ei = GET_EXP_OLD(d,s);
2774: PUT_EXP(r,s,ei);
2775: }
1.1 noro 2776: }
1.43 noro 2777: } else {
1.56 noro 2778: for ( i = 0; i < n; i++ ) {
2779: ei = GET_EXP_OLD(d,i);
2780: PUT_EXP(r,i,ei);
2781: }
1.1 noro 2782: }
2783: }
1.3 noro 2784:
1.6 noro 2785: ND nd_copy(ND p)
2786: {
2787: NM m,mr,mr0;
1.41 noro 2788: int c,n;
1.6 noro 2789: ND r;
2790:
2791: if ( !p )
2792: return 0;
2793: else {
2794: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2795: NEXTNM(mr0,mr);
1.14 noro 2796: CM(mr) = CM(m);
2797: ndl_copy(DL(m),DL(mr));
1.6 noro 2798: }
2799: NEXT(mr) = 0;
1.31 noro 2800: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2801: SG(r) = SG(p);
1.6 noro 2802: return r;
2803: }
2804: }
2805:
1.53 noro 2806: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2807: {
2808: NM m;
2809: NDV p1,p2;
2810: ND t1,t2;
1.61 noro 2811: UINT *lcm;
1.31 noro 2812: int td;
1.11 noro 2813:
1.77 noro 2814: if ( !mod && nd_demand ) {
2815: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 2816: } else {
1.77 noro 2817: if ( trace ) {
2818: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
2819: } else {
2820: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
2821: }
1.20 noro 2822: }
1.34 noro 2823: lcm = LCM(p);
1.11 noro 2824: NEWNM(m);
1.20 noro 2825: CQ(m) = HCQ(p2);
1.34 noro 2826: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 2827: if ( ndl_check_bound2(p->i1,DL(m)) )
2828: return 0;
1.55 noro 2829: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 2830: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
2831: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 2832: else chsgnq(HCQ(p1),&CQ(m));
2833: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 2834: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 2835: nd_free(t1);
2836: return 0;
2837: }
1.55 noro 2838: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 2839: *rp = nd_add(mod,t1,t2);
1.11 noro 2840: FREENM(m);
2841: return 1;
2842: }
2843:
1.19 noro 2844: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 2845: {
2846: NMV m;
2847: int c,c1,len,i;
2848:
1.34 noro 2849: if ( !p ) return;
1.14 noro 2850: len = LEN(p);
1.71 noro 2851: if ( mod == -1 )
2852: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
2853: CM(m) = _mulsf(CM(m),mul);
2854: else
2855: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2856: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2857: }
1.11 noro 2858: }
2859:
1.16 noro 2860: void ndv_mul_c_q(NDV p,Q mul)
2861: {
2862: NMV m;
2863: Q c;
2864: int len,i;
2865:
1.34 noro 2866: if ( !p ) return;
1.16 noro 2867: len = LEN(p);
2868: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2869: mulq(CQ(m),mul,&c); CQ(m) = c;
2870: }
2871: }
2872:
1.55 noro 2873: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
2874: int n2,i,j,l,n,tlen;
1.61 noro 2875: UINT *d0;
1.55 noro 2876: NM *tab,*psum;
2877: ND s,r;
2878: NM t;
2879: NMV m1;
2880:
2881: if ( !p ) return 0;
2882: n = NV(p); n2 = n>>1;
2883: d0 = DL(m0);
2884: l = LEN(p);
2885: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
2886: tab = (NM *)ALLOCA(tlen*sizeof(NM));
2887: psum = (NM *)ALLOCA(tlen*sizeof(NM));
2888: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 2889: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
2890: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 2891: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 2892: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 2893: for ( j = 0; j < tlen; j++ ) {
2894: if ( tab[j] ) {
2895: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
2896: }
2897: }
2898: }
2899: for ( i = tlen-1, r = 0; i >= 0; i-- )
2900: if ( psum[i] ) {
2901: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
2902: MKND(n,psum[i],j,s);
2903: r = nd_add(mod,r,s);
2904: }
1.56 noro 2905: if ( r ) SG(r) = SG(p)+TD(d0);
2906: return r;
1.55 noro 2907: }
2908:
1.56 noro 2909: /* product of monomials */
2910: /* XXX block order is not handled correctly */
2911:
1.55 noro 2912: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
2913: {
1.56 noro 2914: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 2915: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 2916: Q *ctab_q;
2917: Q q,q1;
1.61 noro 2918: UINT c0,c1,c;
1.55 noro 2919: NM *p;
2920: NM m,t;
2921:
2922: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
2923: if ( !m0 || !m1 ) return;
2924: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
2925: NEWNM(m); d = DL(m);
1.56 noro 2926: if ( mod ) {
2927: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
2928: } else
2929: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 2930: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
2931: homo = n&1 ? 1 : 0;
2932: if ( homo ) {
2933: /* offset of h-degree */
2934: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
2935: PUT_EXP(DL(m),n-1,h);
2936: TD(DL(m)) = h;
1.61 noro 2937: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 2938: }
2939: tab[0] = m;
2940: NEWNM(m); d = DL(m);
1.57 noro 2941: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 2942: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
2943: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
2944: /* xi^a*(Di^k*xi^l)*Di^b */
2945: a += l; b += k;
1.56 noro 2946: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 2947: if ( !k || !l ) {
2948: for ( j = 0; j < curlen; j++ )
1.56 noro 2949: if ( t = tab[j] ) {
2950: dt = DL(t);
2951: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 2952: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 2953: }
2954: curlen *= k+1;
2955: continue;
2956: }
2957: min = MIN(k,l);
1.56 noro 2958: if ( mod ) {
1.61 noro 2959: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 2960: mkwcm(k,l,mod,ctab);
2961: } else {
2962: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
2963: mkwc(k,l,ctab_q);
2964: }
1.57 noro 2965: for ( j = min; j >= 0; j-- ) {
1.56 noro 2966: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 2967: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 2968: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 2969: if ( homo ) {
2970: TD(d) = s;
1.56 noro 2971: PUT_EXP(d,n-1,s-h);
1.55 noro 2972: } else TD(d) = h;
1.61 noro 2973: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 2974: if ( mod ) c = ctab[j];
2975: else q = ctab_q[j];
1.57 noro 2976: p = tab+curlen*j;
2977: if ( j == 0 ) {
2978: for ( u = 0; u < curlen; u++, p++ ) {
2979: if ( tab[u] ) {
2980: ndl_addto(DL(tab[u]),d);
2981: if ( mod ) {
2982: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
2983: } else {
2984: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
2985: }
2986: }
1.56 noro 2987: }
1.57 noro 2988: } else {
2989: for ( u = 0; u < curlen; u++, p++ ) {
2990: if ( tab[u] ) {
2991: NEWNM(t);
2992: ndl_add(DL(tab[u]),d,DL(t));
2993: if ( mod ) {
2994: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
2995: } else
2996: mulq(CQ(tab[u]),q,&CQ(t));
2997: *p = t;
2998: }
1.55 noro 2999: }
3000: }
3001: }
3002: curlen *= k+1;
3003: }
3004: FREENM(m);
3005: }
3006:
1.63 noro 3007: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3008: {
3009: NM mr,mr0;
3010: NMV m;
3011: UINT *d,*dt,*dm;
3012: int c,n,td,i,c1,c2,len;
3013: Q q;
3014: ND r;
3015:
3016: if ( !p ) return 0;
3017: else {
3018: n = NV(p); m = BDY(p);
3019: d = DL(m0);
3020: len = LEN(p);
3021: mr0 = 0;
3022: td = TD(d);
3023: c = CM(m0);
3024: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3025: NEXTNM(mr0,mr);
3026: CM(mr) = 1;
3027: ndl_add(DL(m),d,DL(mr));
3028: }
3029: NEXT(mr) = 0;
3030: MKND(NV(p),mr0,len,r);
3031: SG(r) = SG(p) + TD(d);
3032: return r;
3033: }
3034: }
3035:
1.55 noro 3036: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3037: {
3038: NM mr,mr0;
3039: NMV m;
1.61 noro 3040: UINT *d,*dt,*dm;
1.9 noro 3041: int c,n,td,i,c1,c2,len;
1.16 noro 3042: Q q;
1.9 noro 3043: ND r;
3044:
1.34 noro 3045: if ( !p ) return 0;
1.55 noro 3046: else if ( do_weyl )
1.71 noro 3047: if ( mod == -1 )
3048: error("ndv_mul_nm : not implemented (weyl)");
3049: else
3050: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3051: else {
3052: n = NV(p); m = BDY(p);
1.34 noro 3053: d = DL(m0);
1.14 noro 3054: len = LEN(p);
1.9 noro 3055: mr0 = 0;
1.34 noro 3056: td = TD(d);
1.71 noro 3057: if ( mod == -1 ) {
3058: c = CM(m0);
3059: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3060: NEXTNM(mr0,mr);
3061: CM(mr) = _mulsf(CM(m),c);
3062: ndl_add(DL(m),d,DL(mr));
3063: }
3064: } else if ( mod ) {
1.16 noro 3065: c = CM(m0);
3066: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3067: NEXTNM(mr0,mr);
3068: c1 = CM(m);
1.19 noro 3069: DMAR(c1,c,0,mod,c2);
1.16 noro 3070: CM(mr) = c2;
3071: ndl_add(DL(m),d,DL(mr));
3072: }
3073: } else {
3074: q = CQ(m0);
3075: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3076: NEXTNM(mr0,mr);
3077: mulq(CQ(m),q,&CQ(mr));
3078: ndl_add(DL(m),d,DL(mr));
3079: }
1.4 noro 3080: }
1.9 noro 3081: NEXT(mr) = 0;
1.31 noro 3082: MKND(NV(p),mr0,len,r);
1.34 noro 3083: SG(r) = SG(p) + TD(d);
1.9 noro 3084: return r;
1.4 noro 3085: }
3086: }
3087:
1.43 noro 3088: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3089: {
1.13 noro 3090: NMV m,mr,mr0,t;
3091: int len,i,k;
1.11 noro 3092:
1.61 noro 3093: if ( !p ) return;
3094: m = BDY(p); len = LEN(p);
3095: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3096: m = (NMV)((char *)mr0+(len-1)*oadv);
3097: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3098: t = (NMV)ALLOCA(nmv_adv);
3099: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3100: CQ(t) = CQ(m);
3101: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3102: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3103: CQ(mr) = CQ(t);
3104: ndl_copy(DL(t),DL(mr));
3105: }
3106: BDY(p) = mr0;
3107: }
3108:
3109: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3110: {
3111: NMV m,mr,mr0;
3112: int len,i;
3113: NDV r;
1.11 noro 3114:
1.61 noro 3115: if ( !p ) return 0;
3116: m = BDY(p); len = LEN(p);
3117: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3118: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3119: ndl_zero(DL(mr));
3120: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3121: CQ(mr) = CQ(m);
1.11 noro 3122: }
1.61 noro 3123: MKNDV(NV(p),mr0,len,r);
3124: SG(r) = SG(p);
3125: return r;
1.11 noro 3126: }
3127:
1.61 noro 3128: /* duplicate p */
3129:
3130: NDV ndv_dup(int mod,NDV p)
1.3 noro 3131: {
3132: NDV d;
1.61 noro 3133: NMV t,m,m0;
1.3 noro 3134: int i,len;
3135:
1.34 noro 3136: if ( !p ) return 0;
1.31 noro 3137: len = LEN(p);
1.34 noro 3138: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3139: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3140: ndl_copy(DL(t),DL(m));
1.16 noro 3141: CQ(m) = CQ(t);
1.3 noro 3142: }
3143: MKNDV(NV(p),m0,len,d);
1.23 noro 3144: SG(d) = SG(p);
3145: return d;
3146: }
3147:
1.63 noro 3148: ND nd_dup(ND p)
3149: {
3150: ND d;
3151: NM t,m,m0;
3152:
3153: if ( !p ) return 0;
3154: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3155: NEXTNM(m0,m);
3156: ndl_copy(DL(t),DL(m));
3157: CQ(m) = CQ(t);
3158: }
3159: if ( m0 ) NEXT(m) = 0;
3160: MKND(NV(p),m0,LEN(p),d);
3161: SG(d) = SG(p);
3162: return d;
3163: }
3164:
1.61 noro 3165: /* XXX if p->len == 0 then it represents 0 */
3166:
3167: void ndv_mod(int mod,NDV p)
3168: {
3169: NMV t,d;
3170: int r;
3171: int i,len,dlen;
1.71 noro 3172: Obj gfs;
1.61 noro 3173:
3174: if ( !p ) return;
3175: len = LEN(p);
3176: dlen = 0;
1.71 noro 3177: if ( mod == -1 )
3178: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3179: simp_ff((Obj)CP(t),&gfs);
3180: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3181: CM(d) = r;
3182: ndl_copy(DL(t),DL(d));
3183: NMV_ADV(d);
3184: dlen++;
3185: }
1.71 noro 3186: else
3187: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3188: r = rem(NM(CQ(t)),mod);
3189: if ( r ) {
3190: if ( SGN(CQ(t)) < 0 )
3191: r = mod-r;
3192: CM(d) = r;
3193: ndl_copy(DL(t),DL(d));
3194: NMV_ADV(d);
3195: dlen++;
3196: }
3197: }
1.61 noro 3198: LEN(p) = dlen;
3199: }
3200:
3201: NDV ptondv(VL vl,VL dvl,P p)
3202: {
3203: ND nd;
3204:
3205: nd = ptond(vl,dvl,p);
3206: return ndtondv(0,nd);
3207: }
3208:
3209: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3210: {
1.61 noro 3211: int n,i,j,k,e;
3212: VL tvl;
3213: V v;
3214: DCP dc;
3215: DCP *w;
3216: ND r,s,t,u;
3217: P x;
3218: int c;
3219: UINT *d;
1.23 noro 3220: NM m,m0;
1.61 noro 3221:
3222: if ( !p )
3223: return 0;
3224: else if ( NUM(p) ) {
3225: NEWNM(m);
3226: ndl_zero(DL(m));
3227: CQ(m) = (Q)p;
3228: NEXT(m) = 0;
3229: MKND(nd_nvar,m,1,r);
3230: SG(r) = 0;
3231: return r;
3232: } else {
3233: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3234: w = (DCP *)ALLOCA(k*sizeof(DCP));
3235: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3236: for ( i = 0, tvl = dvl, v = VR(p);
3237: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3238: if ( !tvl ) {
3239: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3240: t = ptond(vl,dvl,COEF(w[j]));
3241: pwrp(vl,x,DEG(w[j]),&p);
3242: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3243: }
3244: return s;
3245: } else {
3246: NEWNM(m0); d = DL(m0);
3247: for ( j = k-1, s = 0; j >= 0; j-- ) {
3248: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3249: TD(d) = MUL_WEIGHT(e,i);
3250: if ( nd_blockmask) ndl_weight_mask(d);
3251: t = ptond(vl,dvl,COEF(w[j]));
3252: for ( m = BDY(t); m; m = NEXT(m) )
3253: ndl_addto(DL(m),d);
3254: SG(t) += TD(d);
3255: s = nd_add(0,s,t);
3256: }
3257: FREENM(m0);
3258: return s;
3259: }
3260: }
3261: }
3262:
3263: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3264: {
3265: VL tvl;
3266: int len,n,j,i,e;
3267: NMV m;
3268: Q q;
3269: P c;
3270: UINT *d;
3271: P s,r,u,t,w;
1.71 noro 3272: GFS gfs;
1.23 noro 3273:
1.34 noro 3274: if ( !p ) return 0;
1.61 noro 3275: else {
3276: len = LEN(p);
3277: n = NV(p);
3278: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3279: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3280: if ( mod == -1 ) {
3281: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3282: } else if ( mod ) {
1.61 noro 3283: STOQ(CM(m),q); c = (P)q;
3284: } else
3285: c = CP(m);
3286: d = DL(m);
3287: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3288: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3289: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3290: }
3291: addp(vl,s,t,&u); s = u;
3292: }
3293: return s;
1.23 noro 3294: }
1.3 noro 3295: }
3296:
1.61 noro 3297: NDV ndtondv(int mod,ND p)
1.11 noro 3298: {
3299: NDV d;
1.61 noro 3300: NMV m,m0;
3301: NM t;
3302: int i,len;
1.11 noro 3303:
1.34 noro 3304: if ( !p ) return 0;
1.61 noro 3305: len = LEN(p);
1.74 noro 3306: if ( mod )
3307: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3308: else
3309: m0 = m = MALLOC(len*nmv_adv);
3310: ndv_alloc += nmv_adv*len;
1.61 noro 3311: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3312: ndl_copy(DL(t),DL(m));
3313: CQ(m) = CQ(t);
1.11 noro 3314: }
1.61 noro 3315: MKNDV(NV(p),m0,len,d);
1.14 noro 3316: SG(d) = SG(p);
1.11 noro 3317: return d;
3318: }
3319:
1.61 noro 3320: ND ndvtond(int mod,NDV p)
1.11 noro 3321: {
1.61 noro 3322: ND d;
3323: NM m,m0;
1.11 noro 3324: NMV t;
1.61 noro 3325: int i,len;
1.11 noro 3326:
1.34 noro 3327: if ( !p ) return 0;
1.11 noro 3328: m0 = 0;
1.61 noro 3329: len = p->len;
3330: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3331: NEXTNM(m0,m);
3332: ndl_copy(DL(t),DL(m));
3333: CQ(m) = CQ(t);
1.11 noro 3334: }
3335: NEXT(m) = 0;
1.61 noro 3336: MKND(NV(p),m0,len,d);
1.14 noro 3337: SG(d) = SG(p);
1.11 noro 3338: return d;
3339: }
3340:
1.3 noro 3341: void ndv_print(NDV p)
3342: {
3343: NMV m;
3344: int i,len;
3345:
1.34 noro 3346: if ( !p ) printf("0\n");
1.3 noro 3347: else {
1.14 noro 3348: len = LEN(p);
1.3 noro 3349: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3350: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3351: else printf("+%d*",CM(m));
1.16 noro 3352: ndl_print(DL(m));
3353: }
3354: printf("\n");
3355: }
3356: }
3357:
3358: void ndv_print_q(NDV p)
3359: {
3360: NMV m;
3361: int i,len;
3362:
1.34 noro 3363: if ( !p ) printf("0\n");
1.16 noro 3364: else {
3365: len = LEN(p);
3366: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3367: printf("+");
1.74 noro 3368: printexpr(CO,(Obj)CQ(m));
1.16 noro 3369: printf("*");
1.14 noro 3370: ndl_print(DL(m));
1.3 noro 3371: }
3372: printf("\n");
3373: }
1.25 noro 3374: }
3375:
1.61 noro 3376: NODE ndv_reducebase(NODE x)
1.27 noro 3377: {
3378: int len,i,j;
3379: NDV *w;
3380: NODE t,t0;
3381:
3382: len = length(x);
3383: w = (NDV *)ALLOCA(len*sizeof(NDV));
3384: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3385: for ( i = 0; i < len; i++ ) {
3386: for ( j = 0; j < i; j++ ) {
3387: if ( w[i] && w[j] )
3388: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3389: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3390: }
3391: }
3392: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3393: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3394: }
3395: NEXT(t) = 0; x = t0;
3396: return x;
1.11 noro 3397: }
1.32 noro 3398:
1.43 noro 3399: /* XXX incomplete */
3400:
1.32 noro 3401: void nd_init_ord(struct order_spec *ord)
3402: {
1.43 noro 3403: switch ( ord->id ) {
1.32 noro 3404: case 0:
1.43 noro 3405: switch ( ord->ord.simple ) {
3406: case 0:
3407: nd_dcomp = 1;
3408: nd_isrlex = 1;
3409: break;
3410: case 1:
3411: nd_dcomp = 1;
3412: nd_isrlex = 0;
3413: break;
3414: case 2:
3415: nd_dcomp = 0;
3416: nd_isrlex = 0;
1.45 noro 3417: ndl_compare_function = ndl_lex_compare;
1.58 noro 3418: break;
3419: case 11:
3420: /* XXX */
3421: nd_dcomp = 0;
3422: nd_isrlex = 1;
3423: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3424: break;
3425: default:
3426: error("nd_gr : unsupported order");
3427: }
1.32 noro 3428: break;
3429: case 1:
1.43 noro 3430: /* XXX */
3431: nd_dcomp = -1;
1.32 noro 3432: nd_isrlex = 0;
1.45 noro 3433: ndl_compare_function = ndl_block_compare;
1.34 noro 3434: break;
1.43 noro 3435: case 2:
3436: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3437: break;
3438: }
1.41 noro 3439: nd_ord = ord;
1.32 noro 3440: }
3441:
1.43 noro 3442: BlockMask nd_create_blockmask(struct order_spec *ord)
3443: {
3444: int n,i,j,s,l;
1.61 noro 3445: UINT *t;
1.43 noro 3446: BlockMask bm;
3447:
3448: if ( !ord->id )
3449: return 0;
3450: n = ord->ord.block.length;
3451: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3452: bm->n = n;
3453: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3454: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3455: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3456: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3457: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3458: l = bm->order_pair[i].length;
3459: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3460: }
3461: return bm;
1.57 noro 3462: }
3463:
3464: EPOS nd_create_epos(struct order_spec *ord)
3465: {
3466: int i,j,l,s,ord_l,ord_o;
3467: EPOS epos;
3468: struct order_pair *op;
3469:
3470: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3471: switch ( ord->id ) {
3472: case 0:
3473: if ( nd_isrlex ) {
3474: for ( i = 0; i < nd_nvar; i++ ) {
3475: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3476: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3477: }
3478: } else {
3479: for ( i = 0; i < nd_nvar; i++ ) {
3480: epos[i].i = nd_exporigin + i/nd_epw;
3481: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3482: }
3483: }
3484: break;
3485: case 1:
3486: /* block order */
3487: l = ord->ord.block.length;
3488: op = ord->ord.block.order_pair;
3489: for ( j = 0, s = 0; j < l; j++ ) {
3490: ord_o = op[j].order;
3491: ord_l = op[j].length;
3492: if ( !ord_o )
3493: for ( i = 0; i < ord_l; i++ ) {
3494: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3495: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3496: }
3497: else
3498: for ( i = 0; i < ord_l; i++ ) {
3499: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3500: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3501: }
3502: s += ord_l;
3503: }
3504: break;
3505: case 2:
3506: error("nd_create_epos : matrix order is not supported yet.");
3507: }
3508: return epos;
1.43 noro 3509: }
1.59 noro 3510:
3511: /* external interface */
3512:
3513: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3514: {
1.61 noro 3515: NODE t,in0,in;
1.59 noro 3516: ND nd,nf;
1.61 noro 3517: NDV ndv;
3518: VL vv,tv;
3519: int stat,nvar,max,e;
1.69 noro 3520: union oNDC dn;
1.59 noro 3521:
3522: pltovl(v,&vv);
1.74 noro 3523: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3524:
3525: /* get the degree bound */
3526: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3527: for ( tv = vv; tv; tv = NEXT(tv) ) {
3528: e = getdeg(tv->v,(P)BDY(t));
3529: max = MAX(e,max);
3530: }
3531: for ( tv = vv; tv; tv = NEXT(tv) ) {
3532: e = getdeg(tv->v,f);
3533: max = MAX(e,max);
3534: }
3535:
1.59 noro 3536: nd_init_ord(ord);
1.61 noro 3537: nd_setup_parameters(nvar,max);
3538:
3539: /* conversion to ndv */
3540: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3541: NEXTNODE(in0,in);
3542: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3543: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3544: }
3545: NEXTNODE(in0,in);
3546: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3547: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3548: NEXT(in) = 0;
3549:
3550: ndv_setup(m,0,in0);
1.59 noro 3551: nd_psn--;
3552: nd_scale=2;
3553: while ( 1 ) {
3554: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3555: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3556: if ( !stat ) {
3557: nd_psn++;
3558: nd_reconstruct(m,0,0);
3559: nd_psn--;
3560: } else
3561: break;
3562: }
1.61 noro 3563: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3564: }
3565:
3566: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3567: {
3568: NM m;
3569: UINT *t,*s;
3570: int i;
3571:
3572: for ( i = 0; i < n; i++ ) r[i] = 0;
3573: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3574: t = DL(m);
3575: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3576: r[i] = CM(m);
3577: }
3578: for ( i = 0; !r[i]; i++ );
3579: return i;
3580: }
3581:
1.74 noro 3582: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3583: {
3584: NMV m;
3585: UINT *t,*s;
3586: int i,j,len;
3587:
3588: for ( i = 0; i < n; i++ ) r[i] = 0;
3589: m = BDY(d);
3590: len = LEN(d);
3591: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3592: t = DL(m);
3593: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3594: r[i] = CM(m);
3595: }
3596: for ( i = 0; !r[i]; i++ );
3597: return i;
3598: }
3599:
1.63 noro 3600: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3601: {
3602: NM m;
3603: NMV mr;
3604: UINT *d,*t,*s;
3605: NDV p;
3606: int i,j,len;
3607:
3608: m = pair->mul;
3609: d = DL(m);
3610: p = nd_ps[pair->index];
3611: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3612: for ( i = 0; i < n; i++ ) r[i] = 0;
3613: len = LEN(p);
3614: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3615: ndl_add(d,DL(mr),t);
3616: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3617: r[i] = CM(mr);
3618: }
3619: for ( i = 0; !r[i]; i++ );
3620: return i;
3621: }
3622:
1.67 noro 3623: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3624: {
3625: NM m;
3626: NMV mr;
3627: UINT *d,*t,*s;
3628: NDV p;
1.67 noro 3629: unsigned char *ivc;
3630: unsigned short *ivs;
1.81 noro 3631: UINT *v,*ivi,*s0v;
1.67 noro 3632: int i,j,len,prev,diff,cdiff;
3633: IndArray r;
1.64 noro 3634:
3635: m = pair->mul;
3636: d = DL(m);
3637: p = nd_ps[pair->index];
3638: len = LEN(p);
3639: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3640: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3641: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3642: ndl_add(d,DL(mr),t);
3643: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3644: v[j] = i;
3645: }
1.81 noro 3646: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 3647: r->head = v[0];
3648: diff = 0;
3649: for ( i = 1; i < len; i++ ) {
3650: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3651: }
3652: if ( diff < 256 ) {
3653: r->width = 1;
3654: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3655: r->index.c = ivc;
3656: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3657: } else if ( diff < 65536 ) {
3658: r->width = 2;
3659: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3660: r->index.s = ivs;
3661: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3662: } else {
3663: r->width = 4;
3664: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3665: r->index.i = ivi;
3666: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3667: }
1.67 noro 3668: return r;
1.64 noro 3669: }
3670:
1.65 noro 3671:
1.76 noro 3672: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3673: {
1.67 noro 3674: int i,j,k,len,pos,prev;
1.66 noro 3675: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3676: IndArray ivect;
3677: unsigned char *ivc;
3678: unsigned short *ivs;
3679: unsigned int *ivi;
1.65 noro 3680: NDV redv;
1.67 noro 3681: NMV mr;
1.65 noro 3682: NODE rp;
1.76 noro 3683: int maxrs;
1.65 noro 3684:
1.76 noro 3685: maxrs = 0;
1.74 noro 3686: for ( i = 0; i < nred; i++ ) {
1.65 noro 3687: ivect = imat[i];
1.67 noro 3688: k = ivect->head; svect[k] %= m;
1.65 noro 3689: if ( c = svect[k] ) {
1.76 noro 3690: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3691: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3692: len = LEN(redv); mr = BDY(redv);
3693: svect[k] = 0; prev = k;
3694: switch ( ivect->width ) {
3695: case 1:
3696: ivc = ivect->index.c;
3697: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3698: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3699: prev = pos;
3700: DMA(c1,c,c2,up,lo);
3701: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3702: } else svect[pos] = lo;
3703: }
3704: break;
3705: case 2:
3706: ivs = ivect->index.s;
3707: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3708: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3709: prev = pos;
3710: DMA(c1,c,c2,up,lo);
3711: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3712: } else svect[pos] = lo;
3713: }
3714: break;
3715: case 4:
3716: ivi = ivect->index.i;
3717: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3718: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3719: prev = pos;
3720: DMA(c1,c,c2,up,lo);
3721: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3722: } else svect[pos] = lo;
3723: }
3724: break;
1.65 noro 3725: }
3726: }
3727: }
1.66 noro 3728: for ( i = 0; i < col; i++ )
3729: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3730: return maxrs;
1.65 noro 3731: }
3732:
1.76 noro 3733: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3734: {
3735: int i,j,k,len,pos,prev;
3736: UINT c,c1,c2,c3,up,lo,dmy;
3737: IndArray ivect;
3738: unsigned char *ivc;
3739: unsigned short *ivs;
3740: unsigned int *ivi;
3741: NDV redv;
3742: NMV mr;
3743: NODE rp;
1.76 noro 3744: int maxrs;
1.72 noro 3745:
1.76 noro 3746: maxrs = 0;
1.74 noro 3747: for ( i = 0; i < nred; i++ ) {
1.72 noro 3748: ivect = imat[i];
3749: k = ivect->head; svect[k] %= m;
3750: if ( c = svect[k] ) {
1.76 noro 3751: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3752: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3753: len = LEN(redv); mr = BDY(redv);
3754: svect[k] = 0; prev = k;
3755: switch ( ivect->width ) {
3756: case 1:
3757: ivc = ivect->index.c;
3758: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3759: pos = prev+ivc[j]; prev = pos;
3760: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3761: }
3762: break;
3763: case 2:
3764: ivs = ivect->index.s;
3765: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3766: pos = prev+ivs[j]; prev = pos;
3767: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3768: }
3769: break;
3770: case 4:
3771: ivi = ivect->index.i;
3772: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3773: pos = prev+ivi[j]; prev = pos;
3774: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3775: }
3776: break;
3777: }
3778: }
3779: }
1.76 noro 3780: return maxrs;
1.72 noro 3781: }
3782:
1.65 noro 3783: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3784: {
3785: int j,k,len;
3786: UINT *p;
3787: UINT c;
3788: NDV r;
3789: NMV mr0,mr;
3790:
3791: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3792: if ( !len ) return 0;
3793: else {
1.74 noro 3794: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
3795: ndv_alloc += nmv_adv*len;
1.65 noro 3796: mr = mr0;
3797: p = s0vect;
3798: for ( j = k = 0; j < col; j++, p += nd_wpd )
3799: if ( !rhead[j] ) {
3800: if ( c = vect[k++] ) {
3801: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3802: }
3803: }
3804: MKNDV(nd_nvar,mr0,len,r);
3805: return r;
3806: }
3807: }
3808:
1.74 noro 3809: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 3810: {
3811: ND_pairs t;
3812: NODE sp0,sp;
3813: int stat;
3814: ND spol;
3815:
3816: for ( t = l; t; t = NEXT(t) ) {
3817: stat = nd_sp(m,0,t,&spol);
3818: if ( !stat ) return 0;
3819: if ( spol ) {
3820: add_pbucket_symbolic(bucket,spol);
3821: }
3822: }
1.68 noro 3823: return 1;
1.65 noro 3824: }
3825:
3826: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
3827: {
3828: NODE rp0,rp;
3829: NM mul,head,s0,s;
1.76 noro 3830: int index,col,i,sugar;
1.65 noro 3831: RHist h;
3832: UINT *s0v,*p;
3833: NM_ind_pair pair;
3834: ND red;
3835:
3836: s0 = 0; rp0 = 0; col = 0;
3837: while ( 1 ) {
3838: head = remove_head_pbucket_symbolic(bucket);
3839: if ( !head ) break;
3840: if ( !s0 ) s0 = head;
3841: else NEXT(s) = head;
3842: s = head;
3843: index = ndl_find_reducer(DL(head));
3844: if ( index >= 0 ) {
3845: h = nd_psh[index];
3846: NEWNM(mul);
3847: ndl_sub(DL(head),DL(h),DL(mul));
3848: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 3849: sugar = TD(DL(mul))+SG(nd_ps[index]);
3850: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 3851: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
3852: add_pbucket_symbolic(bucket,nd_remove_head(red));
3853: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
3854: }
3855: col++;
3856: }
1.72 noro 3857: if ( rp0 ) NEXT(rp) = 0;
3858: NEXT(s) = 0;
1.65 noro 3859: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
3860: for ( i = 0, p = s0v, s = s0; i < col;
3861: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
3862: *s0vect = s0v;
3863: *r = rp0;
3864: return col;
3865: }
3866:
1.69 noro 3867: NODE nd_f4(int m)
3868: {
3869: int i,nh,stat,index;
3870: NODE r,g;
3871: ND_pairs d,l,t;
3872: ND spol,red;
3873: NDV nf,redv;
3874: NM s0,s;
1.84 noro 3875: NODE rp0,srp0,nflist;
1.69 noro 3876: int nsp,nred,col,rank,len,k,j,a;
3877: UINT c;
1.74 noro 3878: UINT **spmat;
1.69 noro 3879: UINT *s0vect,*svect,*p,*v;
3880: int *colstat;
3881: IndArray *imat;
3882: int *rhead;
3883: int spcol,sprow;
3884: int sugar;
3885: PGeoBucket bucket;
3886: struct oEGT eg0,eg1,eg_f4;
3887:
3888: if ( !m )
3889: error("nd_f4 : not implemented");
1.74 noro 3890: ndv_alloc = 0;
1.69 noro 3891: g = 0; d = 0;
3892: for ( i = 0; i < nd_psn; i++ ) {
3893: d = update_pairs(d,g,i);
3894: g = update_base(g,i);
3895: }
3896: while ( d ) {
3897: get_eg(&eg0);
3898: l = nd_minsugarp(d,&d);
3899: sugar = SG(l);
3900: bucket = create_pbucket();
1.74 noro 3901: stat = nd_sp_f4(m,l,bucket);
1.69 noro 3902: if ( !stat ) {
3903: for ( t = l; NEXT(t); t = NEXT(t) );
3904: NEXT(t) = d; d = l;
3905: d = nd_reconstruct(m,0,d);
3906: continue;
3907: }
1.86 noro 3908: if ( bucket->m < 0 ) continue;
1.69 noro 3909: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
3910: if ( !col ) {
3911: for ( t = l; NEXT(t); t = NEXT(t) );
3912: NEXT(t) = d; d = l;
3913: d = nd_reconstruct(m,0,d);
3914: continue;
3915: }
3916: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 3917: if ( DP_Print )
3918: fprintf(asir_out,"sugar=%d,symb=%fsec,",
3919: sugar,eg_f4.exectime+eg_f4.gctime);
3920: if ( 1 )
3921: nflist = nd_f4_red(m,l,s0vect,col,rp0);
3922: else
3923: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 3924: /* adding new bases */
1.74 noro 3925: for ( r = nflist; r; r = NEXT(r) ) {
3926: nf = (NDV)BDY(r);
1.69 noro 3927: ndv_removecont(m,nf);
1.77 noro 3928: nh = ndv_newps(m,nf,0);
1.69 noro 3929: d = update_pairs(d,g,nh);
3930: g = update_base(g,nh);
3931: }
3932: }
3933: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 3934: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 3935: return g;
3936: }
1.74 noro 3937:
3938: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 3939: {
1.67 noro 3940: IndArray *imat;
1.74 noro 3941: int nsp,nred,spcol,sprow,a;
1.65 noro 3942: int *rhead;
1.74 noro 3943: int i,j,k,l,rank;
3944: NODE rp,r0,r;
3945: ND_pairs sp;
3946: ND spol;
3947: int **spmat;
3948: UINT *svect,*v;
3949: int *colstat;
1.80 noro 3950: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
1.74 noro 3951: NM_ind_pair *rvect;
1.76 noro 3952: int maxrs;
3953: int *spsugar;
1.63 noro 3954:
1.74 noro 3955: get_eg(&eg0);
3956: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
3957: nred = length(rp0); spcol = col-nred;
3958: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
3959: rhead = (int *)ALLOCA(col*sizeof(int));
3960: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 3961:
1.74 noro 3962: /* construction of index arrays */
3963: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
3964: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
3965: rvect[i] = (NM_ind_pair)BDY(rp);
3966: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
3967: rhead[imat[i]->head] = 1;
3968: }
3969:
3970: /* elimination (1st step) */
3971: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
3972: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 3973: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 3974: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
3975: nd_sp(m,0,sp,&spol);
1.75 noro 3976: if ( !spol ) continue;
1.74 noro 3977: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 3978: if ( m == -1 )
3979: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
3980: else
3981: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 3982: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
3983: if ( i < col ) {
3984: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
3985: for ( j = k = 0; j < col; j++ )
3986: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 3987: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 3988: sprow++;
3989: }
1.76 noro 3990: nd_free(spol);
1.74 noro 3991: }
1.80 noro 3992: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
3993: if ( DP_Print ) {
3994: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
3995: fflush(asir_out);
3996: }
1.74 noro 3997: /* free index arrays */
3998: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
3999:
4000: /* elimination (2nd step) */
4001: colstat = (int *)ALLOCA(spcol*sizeof(int));
4002: if ( m == -1 )
1.76 noro 4003: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4004: else
1.76 noro 4005: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4006: r0 = 0;
4007: for ( i = 0; i < rank; i++ ) {
4008: NEXTNODE(r0,r); BDY(r) =
4009: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4010: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4011: GC_free(spmat[i]);
4012: }
4013: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4014: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4015: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4016: if ( DP_Print ) {
1.80 noro 4017: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4018: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4019: nsp,nred,sprow,spcol,rank);
4020: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4021: }
1.74 noro 4022: return r0;
4023: }
4024:
4025: FILE *nd_write,*nd_read;
4026:
4027: void nd_send_int(int a) {
4028: write_int(nd_write,&a);
4029: }
4030:
4031: void nd_send_intarray(int *p,int len) {
4032: write_intarray(nd_write,p,len);
4033: }
4034:
4035: int nd_recv_int() {
4036: int a;
4037:
4038: read_int(nd_read,&a);
4039: return a;
4040: }
4041:
4042: void nd_recv_intarray(int *p,int len) {
4043: read_intarray(nd_read,p,len);
4044: }
4045:
4046: void nd_send_ndv(NDV p) {
4047: int len,i;
4048: NMV m;
4049:
4050: if ( !p ) nd_send_int(0);
4051: else {
4052: len = LEN(p);
4053: nd_send_int(len);
4054: m = BDY(p);
4055: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4056: nd_send_int(CM(m));
4057: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4058: }
1.74 noro 4059: }
4060: }
4061:
4062: void nd_send_nd(ND p) {
4063: int len,i;
4064: NM m;
4065:
4066: if ( !p ) nd_send_int(0);
4067: else {
4068: len = LEN(p);
4069: nd_send_int(len);
4070: m = BDY(p);
4071: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4072: nd_send_int(CM(m));
4073: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4074: }
1.74 noro 4075: }
4076: }
1.65 noro 4077:
1.74 noro 4078: NDV nd_recv_ndv()
4079: {
4080: int len,i;
4081: NMV m,m0;
4082: NDV r;
1.65 noro 4083:
1.74 noro 4084: len = nd_recv_int();
4085: if ( !len ) return 0;
4086: else {
4087: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4088: ndv_alloc += len*nmv_adv;
4089: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4090: CM(m) = nd_recv_int();
4091: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4092: }
1.74 noro 4093: MKNDV(nd_nvar,m0,len,r);
4094: return r;
4095: }
4096: }
1.65 noro 4097:
1.74 noro 4098: int ox_exec_f4_red(Q proc)
4099: {
4100: Obj obj;
4101: STRING fname;
4102: NODE arg;
4103: int s;
4104: extern int ox_need_conv,ox_file_io;
4105:
4106: MKSTR(fname,"nd_exec_f4_red");
4107: arg = mknode(2,proc,fname);
4108: Pox_cmo_rpc(arg,&obj);
4109: s = get_ox_server_id(QTOS(proc));
4110: nd_write = iofp[s].out;
4111: nd_read = iofp[s].in;
4112: ox_need_conv = ox_file_io = 0;
4113: return s;
4114: }
4115:
4116: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4117: {
4118: int nsp,nred;
4119: int i,rank,s;
4120: NODE rp,r0,r;
4121: ND_pairs sp;
4122: NM_ind_pair pair;
4123: NMV nmv;
4124: NM nm;
4125: NDV nf;
4126: Obj proc,dmy;
4127:
4128: ox_launch_main(0,0,&proc);
4129: s = ox_exec_f4_red((Q)proc);
4130:
4131: nd_send_int(m);
4132: nd_send_int(nd_nvar);
4133: nd_send_int(nd_bpe);
4134: nd_send_int(nd_wpd);
4135: nd_send_int(nmv_adv);
4136:
1.89 noro 4137: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
1.74 noro 4138:
4139: nd_send_int(nd_psn);
4140: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4141:
4142: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4143: nd_send_int(nsp);
4144: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4145: nd_send_int(sp->i1); nd_send_int(sp->i2);
4146: }
4147:
4148: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4149:
4150: nred = length(rp0); nd_send_int(nred);
4151: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4152: pair = (NM_ind_pair)BDY(rp);
4153: nd_send_int(pair->index);
4154: nd_send_intarray(pair->mul->dl,nd_wpd);
4155: }
4156: fflush(nd_write);
4157: rank = nd_recv_int();
4158: fprintf(asir_out,"rank=%d\n",rank);
4159: r0 = 0;
4160: for ( i = 0; i < rank; i++ ) {
4161: nf = nd_recv_ndv();
4162: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4163: }
4164: Pox_shutdown(mknode(1,proc),&dmy);
4165: return r0;
4166: }
4167:
4168: /* server side */
4169:
4170: void nd_exec_f4_red_dist()
4171: {
4172: int m,i,nsp,col,s0size,nred,spcol,j,k;
4173: NM_ind_pair *rp0;
4174: NDV nf;
4175: UINT *s0vect;
4176: IndArray *imat;
4177: int *rhead;
4178: int **spmat;
4179: UINT *svect,*v;
4180: ND_pairs *sp0;
4181: int *colstat;
4182: int a,sprow,rank;
1.89 noro 4183: struct order_spec *ord;
1.74 noro 4184: Obj ordspec;
4185: ND spol;
1.76 noro 4186: int maxrs;
4187: int *spsugar;
1.74 noro 4188:
4189: nd_read = iofp[0].in;
4190: nd_write = iofp[0].out;
4191: m = nd_recv_int();
4192: nd_nvar = nd_recv_int();
4193: nd_bpe = nd_recv_int();
4194: nd_wpd = nd_recv_int();
4195: nmv_adv = nd_recv_int();
4196:
4197: loadobj(nd_read,&ordspec);
1.89 noro 4198: create_order_spec(0,ordspec,&ord);
4199: nd_init_ord(ord);
1.74 noro 4200: nd_setup_parameters(nd_nvar,0);
4201:
4202: nd_psn = nd_recv_int();
4203: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4204: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4205: for ( i = 0; i < nd_psn; i++ ) {
4206: nd_ps[i] = nd_recv_ndv();
4207: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4208: }
4209:
4210: nsp = nd_recv_int();
4211: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4212: for ( i = 0; i < nsp; i++ ) {
4213: NEWND_pairs(sp0[i]);
4214: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4215: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4216: }
4217:
4218: col = nd_recv_int();
4219: s0size = col*nd_wpd;
4220: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4221: nd_recv_intarray(s0vect,s0size);
4222:
4223: nred = nd_recv_int();
4224: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4225: for ( i = 0; i < nred; i++ ) {
4226: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4227: rp0[i]->index = nd_recv_int();
4228: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4229: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4230: }
4231:
4232: spcol = col-nred;
4233: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4234: rhead = (int *)MALLOC(col*sizeof(int));
4235: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4236:
4237: /* construction of index arrays */
4238: for ( i = 0; i < nred; i++ ) {
4239: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4240: rhead[imat[i]->head] = 1;
4241: }
4242:
4243: /* elimination (1st step) */
4244: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4245: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4246: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4247: for ( a = sprow = 0; a < nsp; a++ ) {
4248: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4249: if ( !spol ) continue;
1.74 noro 4250: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4251: if ( m == -1 )
4252: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4253: else
4254: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4255: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4256: if ( i < col ) {
4257: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4258: for ( j = k = 0; j < col; j++ )
4259: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4260: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4261: sprow++;
4262: }
1.76 noro 4263: nd_free(spol);
1.74 noro 4264: }
4265: /* elimination (2nd step) */
4266: colstat = (int *)ALLOCA(spcol*sizeof(int));
4267: if ( m == -1 )
1.76 noro 4268: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4269: else
1.76 noro 4270: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4271: nd_send_int(rank);
4272: for ( i = 0; i < rank; i++ ) {
4273: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4274: nd_send_ndv(nf);
1.63 noro 4275: }
1.74 noro 4276: fflush(nd_write);
1.76 noro 4277: }
4278:
4279: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4280: {
4281: int i,j,k,l,inv,a,rank,s;
4282: unsigned int *t,*pivot,*pk;
4283: unsigned int **mat;
4284:
4285: mat = (unsigned int **)mat0;
4286: for ( rank = 0, j = 0; j < col; j++ ) {
4287: for ( i = rank; i < row; i++ )
4288: mat[i][j] %= md;
4289: for ( i = rank; i < row; i++ )
4290: if ( mat[i][j] )
4291: break;
4292: if ( i == row ) {
4293: colstat[j] = 0;
4294: continue;
4295: } else
4296: colstat[j] = 1;
4297: if ( i != rank ) {
4298: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4299: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4300: }
4301: pivot = mat[rank];
4302: s = sugar[rank];
4303: inv = invm(pivot[j],md);
4304: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4305: if ( *pk ) {
4306: if ( *pk >= (unsigned int)md )
4307: *pk %= md;
4308: DMAR(*pk,inv,0,md,*pk)
4309: }
4310: for ( i = rank+1; i < row; i++ ) {
4311: t = mat[i];
4312: if ( a = t[j] ) {
4313: sugar[i] = MAX(sugar[i],s);
4314: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4315: }
4316: }
4317: rank++;
4318: }
4319: for ( j = col-1, l = rank-1; j >= 0; j-- )
4320: if ( colstat[j] ) {
4321: pivot = mat[l];
4322: s = sugar[l];
4323: for ( i = 0; i < l; i++ ) {
4324: t = mat[i];
4325: t[j] %= md;
4326: if ( a = t[j] ) {
4327: sugar[i] = MAX(sugar[i],s);
4328: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4329: }
4330: }
4331: l--;
4332: }
4333: for ( j = 0, l = 0; l < rank; j++ )
4334: if ( colstat[j] ) {
4335: t = mat[l];
4336: for ( k = j; k < col; k++ )
4337: if ( t[k] >= (unsigned int)md )
4338: t[k] %= md;
4339: l++;
4340: }
4341: return rank;
4342: }
4343:
4344: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4345: {
4346: int i,j,k,l,inv,a,rank,s;
4347: unsigned int *t,*pivot,*pk;
4348: unsigned int **mat;
4349:
4350: mat = (unsigned int **)mat0;
4351: for ( rank = 0, j = 0; j < col; j++ ) {
4352: for ( i = rank; i < row; i++ )
4353: if ( mat[i][j] )
4354: break;
4355: if ( i == row ) {
4356: colstat[j] = 0;
4357: continue;
4358: } else
4359: colstat[j] = 1;
4360: if ( i != rank ) {
4361: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4362: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4363: }
4364: pivot = mat[rank];
4365: s = sugar[rank];
4366: inv = _invsf(pivot[j]);
4367: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4368: if ( *pk )
4369: *pk = _mulsf(*pk,inv);
4370: for ( i = rank+1; i < row; i++ ) {
4371: t = mat[i];
4372: if ( a = t[j] ) {
4373: sugar[i] = MAX(sugar[i],s);
4374: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4375: }
4376: }
4377: rank++;
4378: }
4379: for ( j = col-1, l = rank-1; j >= 0; j-- )
4380: if ( colstat[j] ) {
4381: pivot = mat[l];
4382: s = sugar[l];
4383: for ( i = 0; i < l; i++ ) {
4384: t = mat[i];
4385: if ( a = t[j] ) {
4386: sugar[i] = MAX(sugar[i],s);
4387: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4388: }
4389: }
4390: l--;
4391: }
4392: return rank;
1.77 noro 4393: }
4394:
4395: int ndv_ishomo(NDV p)
4396: {
4397: NMV m;
4398: int len,h;
4399:
4400: if ( !p ) return 1;
4401: len = LEN(p);
4402: m = BDY(p);
4403: h = TD(DL(m));
4404: NMV_ADV(m);
4405: for ( len--; len; len--, NMV_ADV(m) )
4406: if ( TD(DL(m)) != h ) return 0;
4407: return 1;
4408: }
4409:
4410: void ndv_save(NDV p,int index)
4411: {
4412: FILE *s;
4413: char name[BUFSIZ];
4414: short id;
4415: int nv,sugar,len,n,i,td,e,j;
4416: NMV m;
4417: unsigned int *dl;
4418:
4419: sprintf(name,"%s/%d",Demand,index);
4420: s = fopen(name,"w");
4421: savevl(s,0);
4422: if ( !p ) {
4423: saveobj(s,0);
4424: return;
4425: }
4426: id = O_DP;
4427: nv = NV(p);
4428: sugar = SG(p);
4429: len = LEN(p);
4430: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
4431: write_int(s,&len);
4432:
4433: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4434: saveobj(s,(Obj)CQ(m));
4435: dl = DL(m);
4436: td = TD(dl);
4437: write_int(s,&td);
4438: for ( j = 0; j < nv; j++ ) {
4439: e = GET_EXP(dl,j);
4440: write_int(s,&e);
4441: }
4442: }
4443: fclose(s);
4444: }
4445:
4446: NDV ndv_load(int index)
4447: {
4448: FILE *s;
4449: char name[BUFSIZ];
4450: short id;
4451: int nv,sugar,len,n,i,td,e,j;
4452: NDV d;
4453: NMV m0,m;
4454: unsigned int *dl;
4455: Obj obj;
4456:
4457: sprintf(name,"%s/%d",Demand,index);
4458: s = fopen(name,"r");
4459: if ( !s ) return 0;
4460:
4461: skipvl(s);
4462: read_short(s,&id);
4463: if ( !id ) return 0;
4464: read_int(s,&nv);
4465: read_int(s,&sugar);
4466: read_int(s,&len);
4467:
4468: m0 = m = MALLOC(len*nmv_adv);
4469: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4470: loadobj(s,&obj); CQ(m) = (Q)obj;
4471: dl = DL(m);
4472: ndl_zero(dl);
4473: read_int(s,&td); TD(dl) = td;
4474: for ( j = 0; j < nv; j++ ) {
4475: read_int(s,&e);
4476: PUT_EXP(dl,j,e);
4477: }
4478: if ( nd_blockmask ) ndl_weight_mask(dl);
4479: }
4480: fclose(s);
4481: MKNDV(nv,m0,len,d);
4482: SG(d) = sugar;
4483: return d;
1.59 noro 4484: }