Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.158
1.158 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.157 2009/01/04 05:44:51 noro Exp $ */
1.2 noro 2:
1.94 noro 3: #include "nd.h"
1.63 noro 4:
1.131 noro 5: int diag_period = 6;
1.61 noro 6: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.94 noro 7: int nd_dcomp;
8: NM _nm_free_list;
9: ND _nd_free_list;
10: ND_pairs _ndp_free_list;
1.150 noro 11: NODE nd_hcf;
1.32 noro 12:
1.146 noro 13: static NODE nd_subst;
14: static VL nd_vc;
1.121 noro 15: static int nd_ntrans;
1.117 noro 16: static int nd_nalg;
1.103 noro 17: #if 0
1.74 noro 18: static int ndv_alloc;
1.103 noro 19: #endif
1.87 noro 20: #if 1
1.69 noro 21: static int nd_f4_nsp=0x7fffffff;
1.87 noro 22: #else
23: static int nd_f4_nsp=50;
24: #endif
1.42 noro 25: static double nd_scale=2;
1.61 noro 26: static UINT **nd_bound;
1.42 noro 27: static struct order_spec *nd_ord;
28: static EPOS nd_epos;
1.43 noro 29: static BlockMask nd_blockmask;
1.42 noro 30: static int nd_nvar;
31: static int nd_isrlex;
32: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 33: static UINT nd_mask[32];
34: static UINT nd_mask0,nd_mask1;
1.42 noro 35:
1.20 noro 36: static NDV *nd_ps;
1.53 noro 37: static NDV *nd_ps_trace;
1.42 noro 38: static RHist *nd_psh;
39: static int nd_psn,nd_pslen;
40: static RHist *nd_red;
1.96 noro 41: static int *nd_work_vector;
42: static int **nd_matrix;
43: static int nd_matrix_len;
1.97 noro 44: static struct weight_or_block *nd_worb;
45: static int nd_worb_len;
1.42 noro 46: static int nd_found,nd_create,nd_notfirst;
47: static int nmv_adv;
1.77 noro 48: static int nd_demand;
1.157 noro 49: static int nd_module,nd_istop,nd_mpos;
1.1 noro 50:
1.119 noro 51: NumberField get_numberfield();
1.114 noro 52: UINT *nd_det_compute_bound(NDV **dm,int n,int j);
53: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d);
1.152 ohara 54: void nd_heu_nezgcdnpz(VL vl,P *pl,int m,int full,P *pr);
1.118 noro 55: int nd_monic(int m,ND *p);
1.129 noro 56: NDV plain_vect_to_ndv_q(Q *mat,int col,UINT *s0vect);
1.157 noro 57: LIST ndvtopl(int mod,VL vl,VL dvl,NDV p,int rank);
58: NDV pltondv(VL vl,VL dvl,LIST p);
59: void pltozpl(LIST l,Q *cont,LIST *pp);
60: void ndl_lcm_nocheck(UINT *d1,unsigned *d2,UINT *d);
1.114 noro 61:
1.149 noro 62: extern int Denominator,DP_Multiple;
63:
1.1 noro 64: void nd_free_private_storage()
65: {
1.157 noro 66: _nm_free_list = 0;
67: _ndp_free_list = 0;
1.71 noro 68: #if 0
1.157 noro 69: GC_gcollect();
1.71 noro 70: #endif
1.1 noro 71: }
72:
73: void _NM_alloc()
74: {
1.157 noro 75: NM p;
76: int i;
1.1 noro 77:
1.157 noro 78: for ( i = 0; i < 1024; i++ ) {
79: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
80: p->next = _nm_free_list; _nm_free_list = p;
81: }
1.1 noro 82: }
83:
84: void _ND_alloc()
85: {
1.157 noro 86: ND p;
87: int i;
1.1 noro 88:
1.157 noro 89: for ( i = 0; i < 1024; i++ ) {
90: p = (ND)GC_malloc(sizeof(struct oND));
91: p->body = (NM)_nd_free_list; _nd_free_list = p;
92: }
1.1 noro 93: }
94:
95: void _NDP_alloc()
96: {
1.157 noro 97: ND_pairs p;
98: int i;
1.1 noro 99:
1.157 noro 100: for ( i = 0; i < 1024; i++ ) {
101: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
102: +(nd_wpd-1)*sizeof(UINT));
103: p->next = _ndp_free_list; _ndp_free_list = p;
104: }
1.1 noro 105: }
106:
1.30 noro 107: INLINE int nd_length(ND p)
1.1 noro 108: {
1.157 noro 109: NM m;
110: int i;
1.1 noro 111:
1.157 noro 112: if ( !p )
113: return 0;
114: else {
115: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
116: return i;
117: }
1.1 noro 118: }
119:
1.61 noro 120: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 121: {
1.157 noro 122: UINT u1,u2;
123: int i,j;
1.1 noro 124:
1.157 noro 125: if ( nd_module && (MPOS(d1) != MPOS(d2)) ) return 0;
126:
127: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 128: #if USE_UNROLL
1.157 noro 129: switch ( nd_bpe ) {
130: case 3:
131: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
132: u1 = d1[i]; u2 = d2[i];
133: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
134: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
135: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
136: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
137: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
138: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
139: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
140: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
141: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
142: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
143: }
144: return 1;
145: break;
146: case 4:
147: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
148: u1 = d1[i]; u2 = d2[i];
149: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
150: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
151: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
152: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
153: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
154: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
155: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
156: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
157: }
158: return 1;
159: break;
160: case 6:
161: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
162: u1 = d1[i]; u2 = d2[i];
163: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
164: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
165: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
166: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
167: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
168: }
169: return 1;
170: break;
171: case 8:
172: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
173: u1 = d1[i]; u2 = d2[i];
174: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
175: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
176: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
177: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
178: }
179: return 1;
180: break;
181: case 16:
182: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
183: u1 = d1[i]; u2 = d2[i];
184: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
185: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
186: }
187: return 1;
188: break;
189: case 32:
190: for ( i = nd_exporigin; i < nd_wpd; i++ )
191: if ( d1[i] < d2[i] ) return 0;
192: return 1;
193: break;
194: default:
195: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
196: u1 = d1[i]; u2 = d2[i];
197: for ( j = 0; j < nd_epw; j++ )
198: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
199: }
200: return 1;
201: }
1.65 noro 202: #else
1.157 noro 203: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
204: u1 = d1[i]; u2 = d2[i];
205: for ( j = 0; j < nd_epw; j++ )
206: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
207: }
208: return 1;
1.65 noro 209: #endif
1.1 noro 210: }
211:
1.61 noro 212: /*
213: * If the current order is a block order,
214: * then the last block is length 1 and contains
215: * the homo variable. Otherwise, the original
216: * order is either 0 or 2.
217: */
218:
219: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 220: {
1.157 noro 221: int w,i,e,n,omask0;
1.61 noro 222:
1.157 noro 223: omask0 = obpe==32?0xffffffff:((1<<obpe)-1);
224: n = nd_nvar-1;
225: ndl_zero(r);
226: for ( i = 0; i < n; i++ ) {
227: e = GET_EXP_OLD(d,i);
228: PUT_EXP(r,i,e);
229: }
230: w = TD(d);
231: PUT_EXP(r,nd_nvar-1,weight-w);
232: TD(r) = weight;
233: if ( nd_blockmask ) ndl_weight_mask(r);
234: if ( nd_module ) MPOS(r) = MPOS(d);
1.61 noro 235: }
236:
237: void ndl_dehomogenize(UINT *d)
238: {
1.157 noro 239: UINT mask;
240: UINT h;
241: int i,bits;
242:
243: if ( nd_blockmask ) {
244: h = GET_EXP(d,nd_nvar-1);
245: XOR_EXP(d,nd_nvar-1,h);
246: TD(d) -= h;
247: ndl_weight_mask(d);
248: } else {
249: if ( nd_isrlex ) {
250: if ( nd_bpe == 32 ) {
251: h = d[nd_exporigin];
252: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
253: d[i-1] = d[i];
254: d[i-1] = 0;
255: TD(d) -= h;
256: } else {
257: bits = nd_epw*nd_bpe;
258: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
259: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
260: for ( i = nd_exporigin; i < nd_wpd; i++ )
261: d[i] = ((d[i]<<nd_bpe)&mask)
262: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
263: TD(d) -= h;
264: }
265: } else {
266: h = GET_EXP(d,nd_nvar-1);
267: XOR_EXP(d,nd_nvar-1,h);
268: TD(d) -= h;
269: }
270: }
1.23 noro 271: }
272:
1.61 noro 273: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 274: {
1.157 noro 275: UINT t1,t2,u,u1,u2;
276: int i,j,l;
277:
278: if ( nd_module && (MPOS(d1) != MPOS(d2)) )
279: error("ndl_lcm : inconsistent monomials");
280: #if USE_UNROLL
281: switch ( nd_bpe ) {
282: case 3:
283: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
284: u1 = d1[i]; u2 = d2[i];
285: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
286: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
287: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
288: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
289: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
290: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
291: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
292: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
293: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
294: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
295: d[i] = u;
296: }
297: break;
298: case 4:
299: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
300: u1 = d1[i]; u2 = d2[i];
301: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
302: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
303: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
304: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
305: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
306: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
307: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
308: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
309: d[i] = u;
310: }
311: break;
312: case 6:
313: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
314: u1 = d1[i]; u2 = d2[i];
315: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
316: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
317: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
318: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
319: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
320: d[i] = u;
321: }
322: break;
323: case 8:
324: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
325: u1 = d1[i]; u2 = d2[i];
326: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
327: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
328: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
329: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
330: d[i] = u;
331: }
332: break;
333: case 16:
334: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
335: u1 = d1[i]; u2 = d2[i];
336: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
337: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
338: d[i] = u;
339: }
340: break;
341: case 32:
342: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
343: u1 = d1[i]; u2 = d2[i];
344: d[i] = u1>u2?u1:u2;
345: }
346: break;
347: default:
348: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
349: u1 = d1[i]; u2 = d2[i];
350: for ( j = 0, u = 0; j < nd_epw; j++ ) {
351: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
352: }
353: d[i] = u;
354: }
355: break;
356: }
357: #else
358: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
359: u1 = d1[i]; u2 = d2[i];
360: for ( j = 0, u = 0; j < nd_epw; j++ ) {
361: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
362: }
363: d[i] = u;
364: }
365: #endif
366: TD(d) = ndl_weight(d);
367: if ( nd_blockmask ) ndl_weight_mask(d);
368: if ( nd_module ) MPOS(d) = MPOS(d1);
369: }
370:
371: void ndl_lcm_nocheck(UINT *d1,unsigned *d2,UINT *d)
372: {
373: UINT t1,t2,u,u1,u2;
374: int i,j,l;
1.1 noro 375:
1.65 noro 376: #if USE_UNROLL
1.157 noro 377: switch ( nd_bpe ) {
378: case 3:
379: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
380: u1 = d1[i]; u2 = d2[i];
381: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
382: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
383: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
384: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
385: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
386: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
387: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
388: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
389: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
390: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
391: d[i] = u;
392: }
393: break;
394: case 4:
395: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
396: u1 = d1[i]; u2 = d2[i];
397: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
398: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
399: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
400: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
401: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
402: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
403: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
404: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
405: d[i] = u;
406: }
407: break;
408: case 6:
409: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
410: u1 = d1[i]; u2 = d2[i];
411: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
412: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
413: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
414: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
415: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
416: d[i] = u;
417: }
418: break;
419: case 8:
420: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
421: u1 = d1[i]; u2 = d2[i];
422: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
423: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
424: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
425: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
426: d[i] = u;
427: }
428: break;
429: case 16:
430: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
431: u1 = d1[i]; u2 = d2[i];
432: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
433: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
434: d[i] = u;
435: }
436: break;
437: case 32:
438: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
439: u1 = d1[i]; u2 = d2[i];
440: d[i] = u1>u2?u1:u2;
441: }
442: break;
443: default:
444: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
445: u1 = d1[i]; u2 = d2[i];
446: for ( j = 0, u = 0; j < nd_epw; j++ ) {
447: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
448: }
449: d[i] = u;
450: }
451: break;
452: }
1.65 noro 453: #else
1.157 noro 454: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
455: u1 = d1[i]; u2 = d2[i];
456: for ( j = 0, u = 0; j < nd_epw; j++ ) {
457: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
458: }
459: d[i] = u;
460: }
1.65 noro 461: #endif
1.157 noro 462: TD(d) = ndl_weight(d);
463: if ( nd_blockmask ) ndl_weight_mask(d);
464: if ( nd_module ) MPOS(d) = MPOS(d1);
1.57 noro 465: }
466:
1.61 noro 467: int ndl_weight(UINT *d)
1.1 noro 468: {
1.157 noro 469: UINT t,u;
470: int i,j;
1.1 noro 471:
1.157 noro 472: if ( current_dl_weight_vector )
473: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
474: u = GET_EXP(d,i);
475: t += MUL_WEIGHT(u,i);
476: }
477: else
478: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
479: u = d[i];
480: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
481: t += (u&nd_mask0);
482: }
483: return t;
1.1 noro 484: }
485:
1.61 noro 486: void ndl_weight_mask(UINT *d)
1.43 noro 487: {
1.157 noro 488: UINT t,u;
489: UINT *mask;
490: int i,j,k,l;
491:
492: l = nd_blockmask->n;
493: for ( k = 0; k < l; k++ ) {
494: mask = nd_blockmask->mask[k];
495: if ( current_dl_weight_vector )
496: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
497: u = GET_EXP_MASK(d,i,mask);
498: t += MUL_WEIGHT(u,i);
499: }
500: else
501: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
502: u = d[i]&mask[i];
503: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
504: t += (u&nd_mask0);
505: }
506: d[k+1] = t;
507: }
1.43 noro 508: }
509:
1.61 noro 510: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 511: {
1.157 noro 512: int i;
1.1 noro 513:
1.157 noro 514: d1 += nd_exporigin;
515: d2 += nd_exporigin;
516: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
517: if ( *d1 > *d2 )
518: return nd_isrlex ? -1 : 1;
519: else if ( *d1 < *d2 )
520: return nd_isrlex ? 1 : -1;
521: return 0;
1.1 noro 522: }
523:
1.61 noro 524: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 525: {
1.157 noro 526: int i,l,j,ord_o,ord_l;
527: struct order_pair *op;
528: UINT t1,t2,m;
529: UINT *mask;
530:
531: l = nd_blockmask->n;
532: op = nd_blockmask->order_pair;
533: for ( j = 0; j < l; j++ ) {
534: mask = nd_blockmask->mask[j];
535: ord_o = op[j].order;
536: if ( ord_o < 2 )
537: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
538: else if ( t1 < t2 ) return -1;
539: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
540: m = mask[i];
541: t1 = d1[i]&m;
542: t2 = d2[i]&m;
543: if ( t1 > t2 )
544: return !ord_o ? -1 : 1;
545: else if ( t1 < t2 )
546: return !ord_o ? 1 : -1;
547: }
548: }
549: return 0;
1.43 noro 550: }
551:
1.96 noro 552: int ndl_matrix_compare(UINT *d1,UINT *d2)
553: {
1.157 noro 554: int i,j,s;
555: int *v;
1.96 noro 556:
1.157 noro 557: for ( j = 0; j < nd_nvar; j++ )
558: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
559: for ( i = 0; i < nd_matrix_len; i++ ) {
560: v = nd_matrix[i];
561: for ( j = 0, s = 0; j < nd_nvar; j++ )
562: s += v[j]*nd_work_vector[j];
563: if ( s > 0 ) return 1;
564: else if ( s < 0 ) return -1;
565: }
566: return 0;
1.96 noro 567: }
568:
1.97 noro 569: int ndl_composite_compare(UINT *d1,UINT *d2)
570: {
1.157 noro 571: int i,j,s,start,end,len,o;
572: int *v;
573: struct sparse_weight *sw;
574:
575: for ( j = 0; j < nd_nvar; j++ )
576: nd_work_vector[j] = GET_EXP(d1,j)-GET_EXP(d2,j);
577: for ( i = 0; i < nd_worb_len; i++ ) {
578: len = nd_worb[i].length;
579: switch ( nd_worb[i].type ) {
580: case IS_DENSE_WEIGHT:
581: v = nd_worb[i].body.dense_weight;
582: for ( j = 0, s = 0; j < len; j++ )
583: s += v[j]*nd_work_vector[j];
584: if ( s > 0 ) return 1;
585: else if ( s < 0 ) return -1;
586: break;
587: case IS_SPARSE_WEIGHT:
588: sw = nd_worb[i].body.sparse_weight;
589: for ( j = 0, s = 0; j < len; j++ )
590: s += sw[j].value*nd_work_vector[sw[j].pos];
591: if ( s > 0 ) return 1;
592: else if ( s < 0 ) return -1;
593: break;
594: case IS_BLOCK:
595: o = nd_worb[i].body.block.order;
596: start = nd_worb[i].body.block.start;
597: switch ( o ) {
598: case 0:
599: end = start+len;
600: for ( j = start, s = 0; j < end; j++ )
601: s += MUL_WEIGHT(nd_work_vector[j],j);
602: if ( s > 0 ) return 1;
603: else if ( s < 0 ) return -1;
604: for ( j = end-1; j >= start; j-- )
605: if ( nd_work_vector[j] < 0 ) return 1;
606: else if ( nd_work_vector[j] > 0 ) return -1;
607: break;
608: case 1:
609: end = start+len;
610: for ( j = start, s = 0; j < end; j++ )
611: s += MUL_WEIGHT(nd_work_vector[j],j);
612: if ( s > 0 ) return 1;
613: else if ( s < 0 ) return -1;
614: for ( j = start; j < end; j++ )
615: if ( nd_work_vector[j] > 0 ) return 1;
616: else if ( nd_work_vector[j] < 0 ) return -1;
617: break;
618: case 2:
619: for ( j = start; j < end; j++ )
620: if ( nd_work_vector[j] > 0 ) return 1;
621: else if ( nd_work_vector[j] < 0 ) return -1;
622: break;
623: }
624: break;
625: }
626: }
627: return 0;
1.97 noro 628: }
629:
1.58 noro 630: /* TDH -> WW -> TD-> RL */
631:
1.61 noro 632: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 633: {
1.157 noro 634: int i,m,e1,e2;
1.58 noro 635:
1.157 noro 636: if ( TD(d1) > TD(d2) ) return 1;
637: else if ( TD(d1) < TD(d2) ) return -1;
638: m = nd_nvar>>1;
639: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
640: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
641: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
642: }
643: if ( e1 > e2 ) return 1;
644: else if ( e1 < e2 ) return -1;
645: return ndl_lex_compare(d1,d2);
646: }
647:
648: int ndl_module_grlex_compare(UINT *d1,UINT *d2)
649: {
650: int i;
651:
652: if ( nd_istop ) {
653: if ( MPOS(d1) < MPOS(d2) ) return 1;
654: else if ( MPOS(d1) > MPOS(d2) ) return -1;
655: }
656: if ( TD(d1) > TD(d2) ) return 1;
657: else if ( TD(d1) < TD(d2) ) return -1;
658: for ( i = nd_nvar-1; i >= 0; i-- )
659: if ( GET_EXP(d1,i) < GET_EXP(d2,i) ) return 1;
660: else if ( GET_EXP(d1,i) > GET_EXP(d2,i) ) return -1;
661: if ( !nd_istop ) {
662: if ( MPOS(d1) < MPOS(d2) ) return 1;
663: else if ( MPOS(d1) > MPOS(d2) ) return -1;
664: }
665: return 0;
666: }
667:
668: int ndl_module_glex_compare(UINT *d1,UINT *d2)
669: {
670: int i;
671:
672: if ( nd_istop ) {
673: if ( MPOS(d1) < MPOS(d2) ) return 1;
674: else if ( MPOS(d1) > MPOS(d2) ) return -1;
675: }
676: if ( TD(d1) > TD(d2) ) return 1;
677: else if ( TD(d1) < TD(d2) ) return -1;
678: for ( i = 0; i < nd_nvar; i++ )
679: if ( GET_EXP(d1,i) > GET_EXP(d2,i) ) return 1;
680: else if ( GET_EXP(d1,i) < GET_EXP(d2,i) ) return -1;
681: if ( !nd_istop ) {
682: if ( MPOS(d1) < MPOS(d2) ) return 1;
683: else if ( MPOS(d1) > MPOS(d2) ) return -1;
684: }
685: return 0;
686: }
687:
688: int ndl_module_lex_compare(UINT *d1,UINT *d2)
689: {
690: int i;
691:
692: if ( nd_istop ) {
693: if ( MPOS(d1) < MPOS(d2) ) return 1;
694: else if ( MPOS(d1) > MPOS(d2) ) return -1;
695: }
696: for ( i = 0; i < nd_nvar; i++ )
697: if ( GET_EXP(d1,i) > GET_EXP(d2,i) ) return 1;
698: else if ( GET_EXP(d1,i) < GET_EXP(d2,i) ) return -1;
699: if ( !nd_istop ) {
700: if ( MPOS(d1) < MPOS(d2) ) return 1;
701: else if ( MPOS(d1) > MPOS(d2) ) return -1;
702: }
703: return 0;
704: }
705:
706: int ndl_module_block_compare(UINT *d1,UINT *d2)
707: {
708: int i,c;
709:
710: if ( nd_istop ) {
711: if ( MPOS(d1) < MPOS(d2) ) return 1;
712: else if ( MPOS(d1) > MPOS(d2) ) return -1;
713: }
714: if ( c = ndl_block_compare(d1,d2) ) return c;
715: if ( !nd_istop ) {
716: if ( MPOS(d1) < MPOS(d2) ) return 1;
717: else if ( MPOS(d1) > MPOS(d2) ) return -1;
718: }
719: return 0;
720: }
721:
722: int ndl_module_matrix_compare(UINT *d1,UINT *d2)
723: {
724: int i,c;
725:
726: if ( nd_istop ) {
727: if ( MPOS(d1) < MPOS(d2) ) return 1;
728: else if ( MPOS(d1) > MPOS(d2) ) return -1;
729: }
730: if ( c = ndl_matrix_compare(d1,d2) ) return c;
731: if ( !nd_istop ) {
732: if ( MPOS(d1) < MPOS(d2) ) return 1;
733: else if ( MPOS(d1) > MPOS(d2) ) return -1;
734: }
735: return 0;
736: }
737:
738: int ndl_module_composite_compare(UINT *d1,UINT *d2)
739: {
740: int i,c;
741:
742: if ( nd_istop ) {
743: if ( MPOS(d1) > MPOS(d2) ) return 1;
744: else if ( MPOS(d1) < MPOS(d2) ) return -1;
745: }
746: if ( c = ndl_composite_compare(d1,d2) ) return c;
747: if ( !nd_istop ) {
748: if ( MPOS(d1) > MPOS(d2) ) return 1;
749: else if ( MPOS(d1) < MPOS(d2) ) return -1;
750: }
751: return 0;
1.58 noro 752: }
753:
1.61 noro 754: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 755: {
1.157 noro 756: int i;
1.1 noro 757:
1.157 noro 758: switch ( nd_wpd ) {
759: case 2:
760: if ( TD(d2) != TD(d1) ) return 0;
761: if ( d2[1] != d1[1] ) return 0;
762: return 1;
763: break;
764: case 3:
765: if ( TD(d2) != TD(d1) ) return 0;
766: if ( d2[1] != d1[1] ) return 0;
767: if ( d2[2] != d1[2] ) return 0;
768: return 1;
769: break;
770: default:
771: for ( i = 0; i < nd_wpd; i++ )
772: if ( *d1++ != *d2++ ) return 0;
773: return 1;
774: break;
775: }
1.1 noro 776: }
777:
1.61 noro 778: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 779: {
1.157 noro 780: int i;
1.6 noro 781:
1.157 noro 782: switch ( nd_wpd ) {
783: case 2:
784: TD(d2) = TD(d1);
785: d2[1] = d1[1];
786: break;
787: case 3:
788: TD(d2) = TD(d1);
789: d2[1] = d1[1];
790: d2[2] = d1[2];
791: break;
792: default:
793: for ( i = 0; i < nd_wpd; i++ )
794: d2[i] = d1[i];
795: break;
796: }
1.6 noro 797: }
798:
1.61 noro 799: INLINE void ndl_zero(UINT *d)
800: {
1.157 noro 801: int i;
802: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
1.61 noro 803: }
804:
805: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 806: {
1.157 noro 807: int i;
1.1 noro 808:
1.43 noro 809: #if 1
1.157 noro 810: switch ( nd_wpd ) {
811: case 2:
812: TD(d) = TD(d1)+TD(d2);
813: d[1] = d1[1]+d2[1];
814: break;
815: case 3:
816: TD(d) = TD(d1)+TD(d2);
817: d[1] = d1[1]+d2[1];
818: d[2] = d1[2]+d2[2];
819: break;
820: default:
821: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
822: break;
823: }
1.43 noro 824: #else
1.157 noro 825: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.43 noro 826: #endif
1.157 noro 827: if ( nd_module ) {
828: if ( MPOS(d1) && MPOS(d2) && (MPOS(d1) != MPOS(d2)) )
829: error("ndl_add : invalid operation");
830: MPOS(d) = MPOS(d1);
831: }
1.6 noro 832: }
833:
1.55 noro 834: /* d1 += d2 */
1.61 noro 835: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 836: {
1.157 noro 837: int i;
1.55 noro 838:
839: #if 1
1.157 noro 840: switch ( nd_wpd ) {
841: case 2:
842: TD(d1) += TD(d2);
843: d1[1] += d2[1];
844: break;
845: case 3:
846: TD(d1) += TD(d2);
847: d1[1] += d2[1];
848: d1[2] += d2[2];
849: break;
850: default:
851: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
852: break;
853: }
1.55 noro 854: #else
1.157 noro 855: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
1.55 noro 856: #endif
1.157 noro 857: if ( nd_module ) {
858: if ( MPOS(d1) && MPOS(d2) && (MPOS(d1) != MPOS(d2)) )
859: error("ndl_addto : invalid operation");
860: MPOS(d1) = MPOS(d2);
861: }
1.55 noro 862: }
863:
1.61 noro 864: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 865: {
1.157 noro 866: int i;
1.6 noro 867:
1.157 noro 868: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 869: }
870:
1.61 noro 871: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 872: {
1.157 noro 873: UINT t1,t2,u,u1,u2;
874: int i,j;
1.1 noro 875:
1.157 noro 876: if ( nd_module && (MPOS(d1) == MPOS(d2)) ) return 0;
1.65 noro 877: #if USE_UNROLL
1.157 noro 878: switch ( nd_bpe ) {
879: case 3:
880: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
881: u1 = d1[i]; u2 = d2[i];
882: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
883: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
884: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
885: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
886: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
887: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
888: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
889: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
890: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
891: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
892: }
893: return 1;
894: break;
895: case 4:
896: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
897: u1 = d1[i]; u2 = d2[i];
898: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
899: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
900: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
901: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
902: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
903: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
904: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
905: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
906: }
907: return 1;
908: break;
909: case 6:
910: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
911: u1 = d1[i]; u2 = d2[i];
912: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
913: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
914: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
915: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
916: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
917: }
918: return 1;
919: break;
920: case 8:
921: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
922: u1 = d1[i]; u2 = d2[i];
923: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
924: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
925: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
926: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
927: }
928: return 1;
929: break;
930: case 16:
931: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
932: u1 = d1[i]; u2 = d2[i];
933: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
934: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
935: }
936: return 1;
937: break;
938: case 32:
939: for ( i = nd_exporigin; i < nd_wpd; i++ )
940: if ( d1[i] && d2[i] ) return 0;
941: return 1;
942: break;
943: default:
944: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
945: u1 = d1[i]; u2 = d2[i];
946: for ( j = 0; j < nd_epw; j++ ) {
947: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
948: u1 >>= nd_bpe; u2 >>= nd_bpe;
949: }
950: }
951: return 1;
952: break;
953: }
1.65 noro 954: #else
1.157 noro 955: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
956: u1 = d1[i]; u2 = d2[i];
957: for ( j = 0; j < nd_epw; j++ ) {
958: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
959: u1 >>= nd_bpe; u2 >>= nd_bpe;
960: }
961: }
962: return 1;
1.65 noro 963: #endif
1.1 noro 964: }
965:
1.114 noro 966: int ndl_check_bound(UINT *d1,UINT *d2)
1.1 noro 967: {
1.157 noro 968: UINT u2;
969: int i,j,ind,k;
1.1 noro 970:
1.157 noro 971: ind = 0;
1.65 noro 972: #if USE_UNROLL
1.157 noro 973: switch ( nd_bpe ) {
974: case 3:
975: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
976: u2 = d2[i];
977: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
978: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
979: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
980: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
981: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
982: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
983: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
984: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
985: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
986: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
987: }
988: return 0;
989: break;
990: case 4:
991: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
992: u2 = d2[i];
993: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
994: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
995: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
996: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
997: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
998: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
999: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1000: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1001: }
1002: return 0;
1003: break;
1004: case 6:
1005: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1006: u2 = d2[i];
1007: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1008: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1009: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1010: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1011: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1012: }
1013: return 0;
1014: break;
1015: case 8:
1016: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1017: u2 = d2[i];
1018: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1019: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1020: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1021: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1022: }
1023: return 0;
1024: break;
1025: case 16:
1026: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1027: u2 = d2[i];
1028: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1029: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1030: }
1031: return 0;
1032: break;
1033: case 32:
1034: for ( i = nd_exporigin; i < nd_wpd; i++ )
1035: if ( d1[i]+d2[i]<d1[i] ) return 1;
1036: return 0;
1037: break;
1038: default:
1039: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1040: u2 = d2[i];
1041: k = (nd_epw-1)*nd_bpe;
1042: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1043: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1044: }
1045: return 0;
1046: break;
1047: }
1.65 noro 1048: #else
1.157 noro 1049: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1050: u2 = d2[i];
1051: k = (nd_epw-1)*nd_bpe;
1052: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1053: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1054: }
1055: return 0;
1.65 noro 1056: #endif
1.1 noro 1057: }
1058:
1.114 noro 1059: int ndl_check_bound2(int index,UINT *d2)
1060: {
1.157 noro 1061: return ndl_check_bound(nd_bound[index],d2);
1.114 noro 1062: }
1063:
1.61 noro 1064: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1065: {
1.157 noro 1066: int i;
1067: int r;
1.1 noro 1068:
1.157 noro 1069: r = 0;
1070: for ( i = 0; i < nd_wpd; i++ )
1071: r = ((r<<16)+d[i])%REDTAB_LEN;
1072: return r;
1.1 noro 1073: }
1074:
1.63 noro 1075: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1076: {
1.157 noro 1077: RHist r;
1078: int d,k,i;
1.1 noro 1079:
1.157 noro 1080: d = ndl_hash_value(dg);
1081: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1082: if ( ndl_equal(dg,DL(r)) ) {
1083: if ( k > 0 ) nd_notfirst++;
1084: nd_found++;
1085: return r->index;
1086: }
1087: }
1088: if ( Reverse )
1089: for ( i = nd_psn-1; i >= 0; i-- ) {
1090: r = nd_psh[i];
1091: if ( ndl_reducible(dg,DL(r)) ) {
1092: nd_create++;
1093: nd_append_red(dg,i);
1094: return i;
1095: }
1096: }
1097: else
1098: for ( i = 0; i < nd_psn; i++ ) {
1099: r = nd_psh[i];
1100: if ( ndl_reducible(dg,DL(r)) ) {
1101: nd_create++;
1102: nd_append_red(dg,i);
1103: return i;
1104: }
1105: }
1106: return -1;
1.1 noro 1107: }
1108:
1.63 noro 1109: ND nd_merge(ND p1,ND p2)
1110: {
1.157 noro 1111: int n,c;
1112: int t,can,td1,td2;
1113: ND r;
1114: NM m1,m2,mr0,mr,s;
1115:
1116: if ( !p1 ) return p2;
1117: else if ( !p2 ) return p1;
1118: else {
1119: can = 0;
1120: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1121: c = DL_COMPARE(DL(m1),DL(m2));
1122: switch ( c ) {
1123: case 0:
1124: s = m1; m1 = NEXT(m1);
1125: can++; NEXTNM2(mr0,mr,s);
1126: s = m2; m2 = NEXT(m2); FREENM(s);
1127: break;
1128: case 1:
1129: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1130: break;
1131: case -1:
1132: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1133: break;
1134: }
1135: }
1136: if ( !mr0 )
1137: if ( m1 ) mr0 = m1;
1138: else if ( m2 ) mr0 = m2;
1139: else return 0;
1140: else if ( m1 ) NEXT(mr) = m1;
1141: else if ( m2 ) NEXT(mr) = m2;
1142: else NEXT(mr) = 0;
1143: BDY(p1) = mr0;
1144: SG(p1) = MAX(SG(p1),SG(p2));
1145: LEN(p1) = LEN(p1)+LEN(p2)-can;
1146: FREEND(p2);
1147: return p1;
1148: }
1.63 noro 1149: }
1150:
1.31 noro 1151: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1152: {
1.157 noro 1153: int n,c;
1154: int t,can,td1,td2;
1155: ND r;
1156: NM m1,m2,mr0,mr,s;
1157:
1158: if ( !p1 ) return p2;
1159: else if ( !p2 ) return p1;
1160: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1161: else if ( !mod ) return nd_add_q(p1,p2);
1162: else {
1163: can = 0;
1164: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1165: c = DL_COMPARE(DL(m1),DL(m2));
1166: switch ( c ) {
1167: case 0:
1168: t = ((CM(m1))+(CM(m2))) - mod;
1169: if ( t < 0 ) t += mod;
1170: s = m1; m1 = NEXT(m1);
1171: if ( t ) {
1172: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1173: } else {
1174: can += 2; FREENM(s);
1175: }
1176: s = m2; m2 = NEXT(m2); FREENM(s);
1177: break;
1178: case 1:
1179: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1180: break;
1181: case -1:
1182: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1183: break;
1184: }
1185: }
1186: if ( !mr0 )
1187: if ( m1 ) mr0 = m1;
1188: else if ( m2 ) mr0 = m2;
1189: else return 0;
1190: else if ( m1 ) NEXT(mr) = m1;
1191: else if ( m2 ) NEXT(mr) = m2;
1192: else NEXT(mr) = 0;
1193: BDY(p1) = mr0;
1194: SG(p1) = MAX(SG(p1),SG(p2));
1195: LEN(p1) = LEN(p1)+LEN(p2)-can;
1196: FREEND(p2);
1197: return p1;
1198: }
1.95 noro 1199: }
1200:
1201: /* XXX on opteron, the inlined manipulation of destructive additon of
1202: * two NM seems to make gcc optimizer get confused, so the part is
1203: * done in a function.
1204: */
1205:
1.113 noro 1206: int nm_destructive_add_q(NM *m1,NM *m2,NM *mr0,NM *mr)
1.95 noro 1207: {
1.157 noro 1208: NM s;
1209: P t;
1210: int can;
1211:
1212: addp(nd_vc,CP(*m1),CP(*m2),&t);
1213: s = *m1; *m1 = NEXT(*m1);
1214: if ( t ) {
1215: can = 1; NEXTNM2(*mr0,*mr,s); CP(*mr) = (t);
1216: } else {
1217: can = 2; FREENM(s);
1218: }
1219: s = *m2; *m2 = NEXT(*m2); FREENM(s);
1220: return can;
1.95 noro 1221: }
1222:
1.113 noro 1223: ND nd_add_q(ND p1,ND p2)
1.95 noro 1224: {
1.157 noro 1225: int n,c,can;
1226: ND r;
1227: NM m1,m2,mr0,mr,s;
1228: P t;
1229:
1230: if ( !p1 ) return p2;
1231: else if ( !p2 ) return p1;
1232: else {
1233: can = 0;
1234: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1235: c = DL_COMPARE(DL(m1),DL(m2));
1236: switch ( c ) {
1237: case 0:
1.95 noro 1238: #if defined(__x86_64__)
1.157 noro 1239: can += nm_destructive_add_q(&m1,&m2,&mr0,&mr);
1.95 noro 1240: #else
1.157 noro 1241: addp(nd_vc,CP(m1),CP(m2),&t);
1242: s = m1; m1 = NEXT(m1);
1243: if ( t ) {
1244: can++; NEXTNM2(mr0,mr,s); CP(mr) = (t);
1245: } else {
1246: can += 2; FREENM(s);
1247: }
1248: s = m2; m2 = NEXT(m2); FREENM(s);
1.95 noro 1249: #endif
1.157 noro 1250: break;
1251: case 1:
1252: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1253: break;
1254: case -1:
1255: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1256: break;
1257: }
1258: }
1259: if ( !mr0 )
1260: if ( m1 ) mr0 = m1;
1261: else if ( m2 ) mr0 = m2;
1262: else return 0;
1263: else if ( m1 ) NEXT(mr) = m1;
1264: else if ( m2 ) NEXT(mr) = m2;
1265: else NEXT(mr) = 0;
1266: BDY(p1) = mr0;
1267: SG(p1) = MAX(SG(p1),SG(p2));
1268: LEN(p1) = LEN(p1)+LEN(p2)-can;
1269: FREEND(p2);
1270: return p1;
1271: }
1.17 noro 1272: }
1273:
1.71 noro 1274: ND nd_add_sf(ND p1,ND p2)
1275: {
1.157 noro 1276: int n,c,can;
1277: ND r;
1278: NM m1,m2,mr0,mr,s;
1279: int t;
1280:
1281: if ( !p1 ) return p2;
1282: else if ( !p2 ) return p1;
1283: else {
1284: can = 0;
1285: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1286: c = DL_COMPARE(DL(m1),DL(m2));
1287: switch ( c ) {
1288: case 0:
1289: t = _addsf(CM(m1),CM(m2));
1290: s = m1; m1 = NEXT(m1);
1291: if ( t ) {
1292: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1293: } else {
1294: can += 2; FREENM(s);
1295: }
1296: s = m2; m2 = NEXT(m2); FREENM(s);
1297: break;
1298: case 1:
1299: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1300: break;
1301: case -1:
1302: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1303: break;
1304: }
1305: }
1306: if ( !mr0 )
1307: if ( m1 ) mr0 = m1;
1308: else if ( m2 ) mr0 = m2;
1309: else return 0;
1310: else if ( m1 ) NEXT(mr) = m1;
1311: else if ( m2 ) NEXT(mr) = m2;
1312: else NEXT(mr) = 0;
1313: BDY(p1) = mr0;
1314: SG(p1) = MAX(SG(p1),SG(p2));
1315: LEN(p1) = LEN(p1)+LEN(p2)-can;
1316: FREEND(p2);
1317: return p1;
1318: }
1.71 noro 1319: }
1320:
1.146 noro 1321: ND nd_reduce2(int mod,ND d,ND g,NDV p,NM mul,NDC dn)
1322: {
1.157 noro 1323: int c,c1,c2;
1324: Q cg,cred,gcd,tq;
1325: P cgp,credp,gcdp;
1326: Obj tr,tr1;
1327:
1328: if ( mod == -1 )
1329: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1330: else if ( mod ) {
1331: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1332: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1333: } else if ( nd_vc ) {
1334: ezgcdpz(nd_vc,HCP(g),HCP(p),&gcdp);
1335: divsp(nd_vc,HCP(g),gcdp,&cgp); divsp(nd_vc,HCP(p),gcdp,&credp);
1336: chsgnp(cgp,&CP(mul));
1337: nd_mul_c_q(d,credp); nd_mul_c_q(g,credp);
1338: if ( dn ) {
1339: mulr(nd_vc,(Obj)dn->r,(Obj)credp,&tr);
1340: reductr(nd_vc,tr,&tr1); dn->r = (R)tr1;
1341: }
1342: } else {
1343: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1344: chsgnq(cg,&CQ(mul));
1345: nd_mul_c_q(d,(P)cred); nd_mul_c_q(g,(P)cred);
1346: if ( dn ) {
1347: mulq(dn->z,cred,&tq); dn->z = tq;
1348: }
1349: }
1350: return nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.146 noro 1351: }
1352:
1.1 noro 1353: /* ret=1 : success, ret=0 : overflow */
1.146 noro 1354: int nd_nf(int mod,ND d,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1355: {
1.157 noro 1356: NM m,mrd,tail;
1357: NM mul;
1358: int n,sugar,psugar,sugar0,stat,index;
1359: int c,c1,c2,dummy;
1360: RHist h;
1361: NDV p,red;
1362: Q cg,cred,gcd,tq,qq;
1363: double hmag;
1364: P tp,tp1;
1365: Obj tr,tr1;
1366:
1367: if ( dn ) {
1368: if ( mod )
1369: dn->m = 1;
1370: else if ( nd_vc )
1371: dn->r = (R)ONE;
1372: else
1373: dn->z = ONE;
1374: }
1375: if ( !g ) {
1376: *rp = d;
1377: return 1;
1378: }
1379: if ( !mod ) hmag = ((double)p_mag(HCP(g)))*nd_scale;
1380:
1381: sugar0 = sugar = SG(g);
1382: n = NV(g);
1383: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1384: if ( d )
1385: for ( tail = BDY(d); NEXT(tail); tail = NEXT(tail) );
1386: for ( ; g; ) {
1387: index = ndl_find_reducer(HDL(g));
1388: if ( index >= 0 ) {
1389: h = nd_psh[index];
1390: ndl_sub(HDL(g),DL(h),DL(mul));
1391: if ( ndl_check_bound2(index,DL(mul)) ) {
1392: nd_free(g); nd_free(d);
1393: return 0;
1394: }
1395: p = nd_demand ? ndv_load(index) : ps[index];
1396: g = nd_reduce2(mod,d,g,p,mul,dn);
1397: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1398: if ( !mod && g && ((double)(p_mag(HCP(g))) > hmag) ) {
1399: tp = HCP(g);
1400: nd_removecont2(d,g);
1401: if ( dn ) {
1402: if ( nd_vc ) {
1403: divsp(nd_vc,tp,HCP(g),&tp1);
1404: divr(nd_vc,(Obj)dn->r,(Obj)tp1,&tr);
1405: reductr(nd_vc,(Obj)tr,&tr1); dn->r = (R)tr1;
1406: } else {
1407: divq((Q)tp,HCQ(g),&qq);
1408: divq(dn->z,qq,&tq); dn->z = tq;
1409: }
1410: }
1411: hmag = ((double)p_mag(HCP(g)))*nd_scale;
1412: }
1413: } else if ( !full ) {
1414: *rp = g;
1415: return 1;
1416: } else {
1417: m = BDY(g);
1418: if ( NEXT(m) ) {
1419: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1420: } else {
1421: FREEND(g); g = 0;
1422: }
1423: if ( d ) {
1424: NEXT(tail)=m; tail=m; LEN(d)++;
1425: } else {
1426: MKND(n,m,1,d); tail = BDY(d);
1427: }
1428: }
1429: }
1430: if ( d ) SG(d) = sugar;
1431: *rp = d;
1432: return 1;
1.1 noro 1433: }
1.28 noro 1434:
1.53 noro 1435: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1436: {
1.157 noro 1437: int hindex,index;
1438: NDV p;
1439: ND u,d,red;
1440: NODE l;
1441: NM mul,m,mrd,tail;
1442: int sugar,psugar,n,h_reducible;
1443: PGeoBucket bucket;
1444: int c,c1,c2;
1445: Q cg,cred,gcd,zzz;
1446: RHist h;
1447: double hmag,gmag;
1448: int count = 0;
1449: int hcount = 0;
1450:
1451: if ( !g ) {
1452: *rp = 0;
1453: return 1;
1454: }
1455: sugar = SG(g);
1456: n = NV(g);
1457: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1458: bucket = create_pbucket();
1459: add_pbucket(mod,bucket,g);
1460: d = 0;
1461: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1462: while ( 1 ) {
1463: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1464: if ( hindex < 0 ) {
1465: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1466: if ( d ) SG(d) = sugar;
1467: *rp = d;
1468: return 1;
1469: }
1470: g = bucket->body[hindex];
1471: index = ndl_find_reducer(HDL(g));
1472: if ( index >= 0 ) {
1473: count++;
1474: if ( !d ) hcount++;
1475: h = nd_psh[index];
1476: ndl_sub(HDL(g),DL(h),DL(mul));
1477: if ( ndl_check_bound2(index,DL(mul)) ) {
1478: nd_free(d);
1479: free_pbucket(bucket);
1480: *rp = 0;
1481: return 0;
1482: }
1483: p = ps[index];
1484: if ( mod == -1 )
1485: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1486: else if ( mod ) {
1487: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1488: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1489: } else {
1490: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1491: chsgnq(cg,&CQ(mul));
1492: nd_mul_c_q(d,(P)cred);
1493: mulq_pbucket(bucket,cred);
1494: g = bucket->body[hindex];
1495: gmag = (double)p_mag((P)HCQ(g));
1496: }
1497: red = ndv_mul_nm(mod,mul,p);
1498: bucket->body[hindex] = nd_remove_head(g);
1499: red = nd_remove_head(red);
1500: add_pbucket(mod,bucket,red);
1501: psugar = SG(p)+TD(DL(mul));
1502: sugar = MAX(sugar,psugar);
1503: if ( !mod && hmag && (gmag > hmag) ) {
1504: g = normalize_pbucket(mod,bucket);
1505: if ( !g ) {
1506: if ( d ) SG(d) = sugar;
1507: *rp = d;
1508: return 1;
1509: }
1510: nd_removecont2(d,g);
1511: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1512: add_pbucket(mod,bucket,g);
1513: }
1514: } else if ( !full ) {
1515: g = normalize_pbucket(mod,bucket);
1516: if ( g ) SG(g) = sugar;
1517: *rp = g;
1518: return 1;
1519: } else {
1520: m = BDY(g);
1521: if ( NEXT(m) ) {
1522: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1523: } else {
1524: FREEND(g); g = 0;
1525: }
1526: bucket->body[hindex] = g;
1527: NEXT(m) = 0;
1528: if ( d ) {
1529: NEXT(tail)=m; tail=m; LEN(d)++;
1530: } else {
1531: MKND(n,m,1,d); tail = BDY(d);
1532: }
1533: }
1534: }
1.25 noro 1535: }
1.27 noro 1536:
1.61 noro 1537: /* input : list of NDV, cand : list of NDV */
1.28 noro 1538:
1.61 noro 1539: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1540: {
1.157 noro 1541: int n,i,stat;
1542: ND nf,d;
1543: NDV r;
1544: NODE t,s;
1545: union oNDC dn;
1.45 noro 1546:
1.157 noro 1547: ndv_setup(0,0,cand,0,1);
1548: n = length(cand);
1.28 noro 1549:
1.157 noro 1550: /* membercheck : list is a subset of Id(cand) ? */
1551: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1552: again:
1.157 noro 1553: if ( nd_bpe > obpe )
1554: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1555: else
1556: r = (NDV)BDY(t);
1557: d = ndvtond(0,r);
1558: stat = nd_nf(0,0,d,nd_ps,0,0,&nf);
1559: if ( !stat ) {
1560: nd_reconstruct(0,0);
1561: goto again;
1562: } else if ( nf ) return 0;
1563: if ( DP_Print ) { printf("."); fflush(stdout); }
1564: }
1565: if ( DP_Print ) { printf("\n"); }
1566: /* gbcheck : cand is a GB of Id(cand) ? */
1567: if ( !nd_gb(0,0,1) ) return 0;
1568: /* XXX */
1569: return 1;
1.23 noro 1570: }
1.1 noro 1571:
1572: ND nd_remove_head(ND p)
1573: {
1.157 noro 1574: NM m;
1.1 noro 1575:
1.157 noro 1576: m = BDY(p);
1577: if ( !NEXT(m) ) {
1578: FREEND(p); p = 0;
1579: } else {
1580: BDY(p) = NEXT(m); LEN(p)--;
1581: }
1582: FREENM(m);
1583: return p;
1.1 noro 1584: }
1585:
1.69 noro 1586: ND nd_separate_head(ND p,ND *head)
1587: {
1.157 noro 1588: NM m,m0;
1589: ND r;
1.69 noro 1590:
1.157 noro 1591: m = BDY(p);
1592: if ( !NEXT(m) ) {
1593: *head = p; p = 0;
1594: } else {
1595: m0 = m;
1596: BDY(p) = NEXT(m); LEN(p)--;
1597: NEXT(m0) = 0;
1598: MKND(NV(p),m0,1,r);
1599: *head = r;
1600: }
1601: return p;
1.69 noro 1602: }
1603:
1.1 noro 1604: PGeoBucket create_pbucket()
1605: {
1606: PGeoBucket g;
1.157 noro 1607:
1608: g = CALLOC(1,sizeof(struct oPGeoBucket));
1609: g->m = -1;
1610: return g;
1.1 noro 1611: }
1612:
1.25 noro 1613: void free_pbucket(PGeoBucket b) {
1.157 noro 1614: int i;
1.25 noro 1615:
1.157 noro 1616: for ( i = 0; i <= b->m; i++ )
1617: if ( b->body[i] ) {
1618: nd_free(b->body[i]);
1619: b->body[i] = 0;
1620: }
1621: GC_free(b);
1.25 noro 1622: }
1623:
1.63 noro 1624: void add_pbucket_symbolic(PGeoBucket g,ND d)
1625: {
1.157 noro 1626: int l,i,k,m;
1.63 noro 1627:
1.157 noro 1628: if ( !d )
1629: return;
1630: l = LEN(d);
1631: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1632: /* 2^(k-1) < l <= 2^k (=m) */
1633: d = nd_merge(g->body[k],d);
1634: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1635: g->body[k] = 0;
1636: d = nd_merge(g->body[k+1],d);
1637: }
1638: g->body[k] = d;
1639: g->m = MAX(g->m,k);
1.63 noro 1640: }
1641:
1.31 noro 1642: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1643: {
1.157 noro 1644: int l,i,k,m;
1.1 noro 1645:
1.157 noro 1646: if ( !d )
1647: return;
1648: l = LEN(d);
1649: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1650: /* 2^(k-1) < l <= 2^k (=m) */
1651: d = nd_add(mod,g->body[k],d);
1652: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1653: g->body[k] = 0;
1654: d = nd_add(mod,g->body[k+1],d);
1655: }
1656: g->body[k] = d;
1657: g->m = MAX(g->m,k);
1.1 noro 1658: }
1659:
1.113 noro 1660: void mulq_pbucket(PGeoBucket g,Q c)
1.26 noro 1661: {
1.157 noro 1662: int k;
1.26 noro 1663:
1.157 noro 1664: for ( k = 0; k <= g->m; k++ )
1665: nd_mul_c_q(g->body[k],(P)c);
1.26 noro 1666: }
1667:
1.63 noro 1668: NM remove_head_pbucket_symbolic(PGeoBucket g)
1669: {
1.157 noro 1670: int j,i,k,c;
1671: NM head;
1672:
1673: k = g->m;
1674: j = -1;
1675: for ( i = 0; i <= k; i++ ) {
1676: if ( !g->body[i] ) continue;
1677: if ( j < 0 ) j = i;
1678: else {
1679: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1680: if ( c > 0 )
1681: j = i;
1682: else if ( c == 0 )
1683: g->body[i] = nd_remove_head(g->body[i]);
1684: }
1685: }
1686: if ( j < 0 ) return 0;
1687: else {
1688: head = BDY(g->body[j]);
1689: if ( !NEXT(head) ) {
1690: FREEND(g->body[j]);
1691: g->body[j] = 0;
1692: } else {
1693: BDY(g->body[j]) = NEXT(head);
1694: LEN(g->body[j])--;
1695: }
1696: return head;
1697: }
1.63 noro 1698: }
1699:
1.19 noro 1700: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1701: {
1.157 noro 1702: int j,i,c,k,nv,sum;
1703: UINT *di,*dj;
1704: ND gi,gj;
1705:
1706: k = g->m;
1707: while ( 1 ) {
1708: j = -1;
1709: for ( i = 0; i <= k; i++ ) {
1710: if ( !(gi = g->body[i]) )
1711: continue;
1712: if ( j < 0 ) {
1713: j = i;
1714: gj = g->body[j];
1715: dj = HDL(gj);
1716: sum = HCM(gj);
1717: } else {
1718: c = DL_COMPARE(HDL(gi),dj);
1719: if ( c > 0 ) {
1720: if ( sum ) HCM(gj) = sum;
1721: else g->body[j] = nd_remove_head(gj);
1722: j = i;
1723: gj = g->body[j];
1724: dj = HDL(gj);
1725: sum = HCM(gj);
1726: } else if ( c == 0 ) {
1727: if ( mod == -1 )
1728: sum = _addsf(sum,HCM(gi));
1729: else {
1730: sum = sum+HCM(gi)-mod;
1731: if ( sum < 0 ) sum += mod;
1732: }
1733: g->body[i] = nd_remove_head(gi);
1734: }
1735: }
1736: }
1737: if ( j < 0 ) return -1;
1738: else if ( sum ) {
1739: HCM(gj) = sum;
1740: return j;
1741: } else
1742: g->body[j] = nd_remove_head(gj);
1743: }
1.26 noro 1744: }
1745:
1.113 noro 1746: int head_pbucket_q(PGeoBucket g)
1.26 noro 1747: {
1.157 noro 1748: int j,i,c,k,nv;
1749: Q sum,t;
1750: ND gi,gj;
1751:
1752: k = g->m;
1753: while ( 1 ) {
1754: j = -1;
1755: for ( i = 0; i <= k; i++ ) {
1756: if ( !(gi = g->body[i]) ) continue;
1757: if ( j < 0 ) {
1758: j = i;
1759: gj = g->body[j];
1760: sum = HCQ(gj);
1761: } else {
1762: nv = NV(gi);
1763: c = DL_COMPARE(HDL(gi),HDL(gj));
1764: if ( c > 0 ) {
1765: if ( sum ) HCQ(gj) = sum;
1766: else g->body[j] = nd_remove_head(gj);
1767: j = i;
1768: gj = g->body[j];
1769: sum = HCQ(gj);
1770: } else if ( c == 0 ) {
1771: addq(sum,HCQ(gi),&t);
1772: sum = t;
1773: g->body[i] = nd_remove_head(gi);
1774: }
1775: }
1776: }
1777: if ( j < 0 ) return -1;
1778: else if ( sum ) {
1779: HCQ(gj) = sum;
1780: return j;
1781: } else
1782: g->body[j] = nd_remove_head(gj);
1783: }
1.1 noro 1784: }
1785:
1.25 noro 1786: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1787: {
1.157 noro 1788: int i;
1789: ND r,t;
1.1 noro 1790:
1.157 noro 1791: r = 0;
1792: for ( i = 0; i <= g->m; i++ ) {
1793: r = nd_add(mod,r,g->body[i]);
1794: g->body[i] = 0;
1795: }
1796: g->m = -1;
1797: return r;
1.1 noro 1798: }
1799:
1.150 noro 1800: #if 0
1801: void register_hcf(NDV p)
1802: {
1.157 noro 1803: DCP dc,t;
1804: P hc,h;
1805: int c;
1806: NODE l,l1,prev;
1807:
1808: hc = p->body->c.p;
1809: if ( !nd_vc || NUM(hc) ) return;
1810: fctrp(nd_vc,hc,&dc);
1811: for ( t = dc; t; t = NEXT(t) ) {
1812: h = t->c;
1813: if ( NUM(h) ) continue;
1814: for ( prev = 0, l = nd_hcf; l; prev = l, l = NEXT(l) ) {
1815: c = compp(nd_vc,h,(P)BDY(l));
1816: if ( c >= 0 ) break;
1817: }
1818: if ( !l || c > 0 ) {
1819: MKNODE(l1,h,l);
1820: if ( !prev )
1821: nd_hcf = l1;
1822: else
1823: NEXT(prev) = l1;
1824: }
1825: }
1.150 noro 1826: }
1827: #else
1828: void register_hcf(NDV p)
1829: {
1.157 noro 1830: DCP dc,t;
1831: P hc,h,q;
1832: Q dmy;
1833: int c;
1834: NODE l,l1,prev;
1835:
1836: hc = p->body->c.p;
1837: if ( NUM(hc) ) return;
1838: ptozp(hc,1,&dmy,&h);
1.150 noro 1839: #if 1
1.157 noro 1840: for ( l = nd_hcf; l; l = NEXT(l) ) {
1841: while ( 1 ) {
1842: if ( divtpz(nd_vc,h,(P)BDY(l),&q) ) h = q;
1843: else break;
1844: }
1845: }
1846: if ( NUM(h) ) return;
1.150 noro 1847: #endif
1.157 noro 1848: for ( prev = 0, l = nd_hcf; l; prev = l, l = NEXT(l) ) {
1849: c = compp(nd_vc,h,(P)BDY(l));
1850: if ( c >= 0 ) break;
1851: }
1852: if ( !l || c > 0 ) {
1853: MKNODE(l1,h,l);
1854: if ( !prev )
1855: nd_hcf = l1;
1856: else
1857: NEXT(prev) = l1;
1858: }
1.150 noro 1859: }
1860: #endif
1861:
1.122 noro 1862: int do_diagonalize(int sugar,int m)
1.92 noro 1863: {
1.157 noro 1864: int i,nh,stat;
1865: NODE r,g,t;
1866: ND h,nf,s,head;
1867: NDV nfv;
1868: Q q,num,den;
1869: P nm,nmp,dn,mnp,dnp;
1870:
1871: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1872: if ( nd_demand )
1873: nfv = ndv_load(i);
1874: else
1875: nfv = nd_ps[i];
1876: s = ndvtond(m,nfv);
1877: s = nd_separate_head(s,&head);
1878: stat = nd_nf(m,head,s,nd_ps,1,0,&nf);
1879: if ( !stat ) return 0;
1880: ndv_free(nfv);
1881: nd_removecont(m,nf);
1882: nfv = ndtondv(m,nf);
1883: nd_free(nf);
1884: nd_bound[i] = ndv_compute_bound(nfv);
1885: if ( !m ) register_hcf(nfv);
1886: if ( nd_demand ) {
1887: ndv_save(nfv,i);
1888: ndv_free(nfv);
1889: } else
1890: nd_ps[i] = nfv;
1891: }
1892: return 1;
1.92 noro 1893: }
1894:
1.27 noro 1895: /* return value = 0 => input is not a GB */
1896:
1.77 noro 1897: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1898: {
1.157 noro 1899: int i,nh,sugar,stat;
1900: NODE r,g,t;
1901: ND_pairs d;
1902: ND_pairs l;
1903: ND h,nf,s,head,nf1;
1904: NDV nfv;
1905: Q q,num,den;
1906: union oNDC dn;
1907: int diag_count = 0;
1908:
1909: g = 0; d = 0;
1910: for ( i = 0; i < nd_psn; i++ ) {
1911: d = update_pairs(d,g,i);
1912: g = update_base(g,i);
1913: }
1914: sugar = 0;
1915: while ( d ) {
1.1 noro 1916: again:
1.157 noro 1917: l = nd_minp(d,&d);
1918: if ( SG(l) != sugar ) {
1919: if ( ishomo ) {
1920: diag_count = 0;
1921: stat = do_diagonalize(sugar,m);
1922: if ( !stat ) {
1923: NEXT(l) = d; d = l;
1924: d = nd_reconstruct(0,d);
1925: goto again;
1926: }
1927: }
1928: sugar = SG(l);
1929: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1930: }
1931: stat = nd_sp(m,0,l,&h);
1932: if ( !stat ) {
1933: NEXT(l) = d; d = l;
1934: d = nd_reconstruct(0,d);
1935: goto again;
1936: }
1.41 noro 1937: #if USE_GEOBUCKET
1.157 noro 1938: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1939: #else
1.157 noro 1940: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 1941: #endif
1.157 noro 1942: if ( !stat ) {
1943: NEXT(l) = d; d = l;
1944: d = nd_reconstruct(0,d);
1945: goto again;
1946: } else if ( nf ) {
1947: if ( checkonly ) return 0;
1948: if ( DP_Print ) { printf("+"); fflush(stdout); }
1949: nd_removecont(m,nf);
1950: if ( !m && nd_nalg ) {
1951: nd_monic(0,&nf);
1952: nd_removecont(m,nf);
1953: }
1954: nfv = ndtondv(m,nf); nd_free(nf);
1955: nh = ndv_newps(m,nfv,0);
1956: if ( !m && (ishomo && ++diag_count == diag_period) ) {
1957: diag_count = 0;
1958: stat = do_diagonalize(sugar,m);
1959: if ( !stat ) {
1960: NEXT(l) = d; d = l;
1961: d = nd_reconstruct(1,d);
1962: goto again;
1963: }
1964: }
1965: d = update_pairs(d,g,nh);
1966: g = update_base(g,nh);
1967: FREENDP(l);
1968: } else {
1969: if ( DP_Print ) { printf("."); fflush(stdout); }
1970: FREENDP(l);
1971: }
1972: }
1973: if ( nd_demand )
1974: for ( t = g; t; t = NEXT(t) )
1975: BDY(t) = (pointer)ndv_load((int)BDY(t));
1976: else
1977: for ( t = g; t; t = NEXT(t) )
1978: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1979: if ( !checkonly && DP_Print ) { printf("nd_gb done.\n"); fflush(stdout); }
1980: return g;
1.1 noro 1981: }
1982:
1.122 noro 1983: int do_diagonalize_trace(int sugar,int m)
1.91 noro 1984: {
1.157 noro 1985: int i,nh,stat;
1986: NODE r,g,t;
1987: ND h,nf,nfq,s,head;
1988: NDV nfv,nfqv;
1989: Q q,den,num;
1990:
1991: for ( i = nd_psn-1; i >= 0 && SG(nd_psh[i]) == sugar; i-- ) {
1992: /* for nd_ps */
1993: s = ndvtond(m,nd_ps[i]);
1994: s = nd_separate_head(s,&head);
1995: stat = nd_nf_pbucket(m,s,nd_ps,1,&nf);
1996: if ( !stat ) return 0;
1997: nf = nd_add(m,head,nf);
1998: ndv_free(nd_ps[i]);
1999: nd_ps[i] = ndtondv(m,nf);
2000: nd_free(nf);
2001:
2002: /* for nd_ps_trace */
2003: if ( nd_demand )
2004: nfv = ndv_load(i);
2005: else
2006: nfv = nd_ps_trace[i];
2007: s = ndvtond(0,nfv);
2008: s = nd_separate_head(s,&head);
2009: stat = nd_nf(0,head,s,nd_ps_trace,1,0,&nf);
2010: if ( !stat ) return 0;
2011: ndv_free(nfv);
2012: nd_removecont(0,nf);
2013: nfv = ndtondv(0,nf);
2014: nd_free(nf);
2015: nd_bound[i] = ndv_compute_bound(nfv);
2016: register_hcf(nfv);
2017: if ( nd_demand ) {
2018: ndv_save(nfv,i);
2019: ndv_free(nfv);
2020: } else
2021: nd_ps_trace[i] = nfv;
2022: }
2023: return 1;
1.91 noro 2024: }
2025:
1.118 noro 2026: static struct oEGT eg_invdalg;
2027: struct oEGT eg_le;
2028:
1.147 noro 2029: void nd_subst_vector(VL vl,P p,NODE subst,P *r)
2030: {
1.157 noro 2031: NODE tn;
2032: P p1;
1.147 noro 2033:
1.157 noro 2034: for ( tn = subst; tn; tn = NEXT(NEXT(tn)) ) {
2035: substp(vl,p,BDY(tn),BDY(NEXT(tn)),&p1); p = p1;
2036: }
2037: *r = p;
1.147 noro 2038: }
2039:
1.82 noro 2040: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 2041: {
1.157 noro 2042: int i,nh,sugar,stat;
2043: NODE r,g,t;
2044: ND_pairs d;
2045: ND_pairs l;
2046: ND h,nf,nfq,s,head;
2047: NDV nfv,nfqv;
2048: Q q,den,num;
2049: P hc;
2050: union oNDC dn;
2051: struct oEGT eg_monic,egm0,egm1;
2052: int diag_count = 0;
2053:
2054: init_eg(&eg_monic);
2055: init_eg(&eg_invdalg);
2056: init_eg(&eg_le);
2057: g = 0; d = 0;
2058: for ( i = 0; i < nd_psn; i++ ) {
2059: d = update_pairs(d,g,i);
2060: g = update_base(g,i);
2061: }
2062: sugar = 0;
2063: while ( d ) {
1.20 noro 2064: again:
1.157 noro 2065: l = nd_minp(d,&d);
2066: if ( SG(l) != sugar ) {
1.130 noro 2067: #if 1
1.157 noro 2068: if ( ishomo ) {
2069: if ( DP_Print > 2 ) fprintf(asir_out,"|");
2070: stat = do_diagonalize_trace(sugar,m);
2071: if ( DP_Print > 2 ) fprintf(asir_out,"|");
2072: diag_count = 0;
2073: if ( !stat ) {
2074: NEXT(l) = d; d = l;
2075: d = nd_reconstruct(1,d);
2076: goto again;
2077: }
2078: }
1.130 noro 2079: #endif
1.157 noro 2080: sugar = SG(l);
2081: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
2082: }
2083: stat = nd_sp(m,0,l,&h);
2084: if ( !stat ) {
2085: NEXT(l) = d; d = l;
2086: d = nd_reconstruct(1,d);
2087: goto again;
2088: }
1.41 noro 2089: #if USE_GEOBUCKET
1.157 noro 2090: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 2091: #else
1.157 noro 2092: stat = nd_nf(m,0,h,nd_ps,!Top,0,&nf);
1.41 noro 2093: #endif
1.157 noro 2094: if ( !stat ) {
2095: NEXT(l) = d; d = l;
2096: d = nd_reconstruct(1,d);
2097: goto again;
2098: } else if ( nf ) {
2099: if ( nd_demand ) {
2100: nfqv = ndv_load(nd_psn);
2101: nfq = ndvtond(0,nfqv);
2102: } else
2103: nfq = 0;
2104: if ( !nfq ) {
2105: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,0,h,nd_ps_trace,!Top,0,&nfq) ) {
2106: NEXT(l) = d; d = l;
2107: d = nd_reconstruct(1,d);
2108: goto again;
2109: }
2110: }
2111: if ( nfq ) {
2112: /* m|HC(nfq) => failure */
2113: if ( nd_vc ) {
2114: nd_subst_vector(nd_vc,HCP(nfq),nd_subst,&hc); q = (Q)hc;
2115: } else
2116: q = HCQ(nfq);
2117: if ( !rem(NM(q),m) ) return 0;
2118:
2119: if ( DP_Print ) { printf("+"); fflush(stdout); }
2120: if ( nd_nalg ) {
2121: /* m|DN(HC(nf)^(-1)) => failure */
2122: get_eg(&egm0);
2123: if ( !nd_monic(m,&nfq) ) return 0;
2124: get_eg(&egm1); add_eg(&eg_monic,&egm0,&egm1);
2125: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
2126: nfv = ndv_dup(0,nfqv); ndv_mod(m,nfv); nd_free(nf);
2127: } else {
2128: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
2129: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
2130: }
2131: nh = ndv_newps(0,nfv,nfqv);
2132: if ( ishomo && ++diag_count == diag_period ) {
2133: diag_count = 0;
2134: if ( DP_Print > 2 ) fprintf(asir_out,"|");
2135: stat = do_diagonalize_trace(sugar,m);
2136: if ( DP_Print > 2 ) fprintf(asir_out,"|");
2137: if ( !stat ) {
2138: NEXT(l) = d; d = l;
2139: d = nd_reconstruct(1,d);
2140: goto again;
2141: }
2142: }
2143: d = update_pairs(d,g,nh);
2144: g = update_base(g,nh);
2145: } else {
2146: if ( DP_Print ) { printf("*"); fflush(stdout); }
2147: }
2148: } else {
2149: if ( DP_Print ) { printf("."); fflush(stdout); }
2150: }
2151: FREENDP(l);
2152: }
2153: if ( nd_demand )
2154: for ( t = g; t; t = NEXT(t) )
2155: BDY(t) = (pointer)ndv_load((int)BDY(t));
2156: else
2157: for ( t = g; t; t = NEXT(t) )
2158: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
2159: if ( nd_nalg ) {
2160: print_eg("monic",&eg_monic);
2161: print_eg("invdalg",&eg_invdalg);
2162: print_eg("le",&eg_le);
2163: }
2164: if ( DP_Print ) { printf("nd_gb_trace done.\n"); fflush(stdout); }
2165: return g;
1.20 noro 2166: }
2167:
1.23 noro 2168: int ndv_compare(NDV *p1,NDV *p2)
2169: {
1.157 noro 2170: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 2171: }
2172:
2173: int ndv_compare_rev(NDV *p1,NDV *p2)
2174: {
1.157 noro 2175: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 2176: }
2177:
1.61 noro 2178: NODE ndv_reduceall(int m,NODE f)
1.23 noro 2179: {
1.157 noro 2180: int i,n,stat;
2181: ND nf,g,head;
2182: NODE t,a0,a;
2183: union oNDC dn;
2184: NDV *w;
2185: Q q,num,den;
1.23 noro 2186:
1.157 noro 2187: n = length(f);
1.69 noro 2188: #if 0
1.157 noro 2189: w = (NDV *)ALLOCA(n*sizeof(NDV));
2190: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
2191: qsort(w,n,sizeof(NDV),
2192: (int (*)(const void *,const void *))ndv_compare);
2193: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1.69 noro 2194: #endif
1.157 noro 2195: ndv_setup(m,0,f,0,1);
2196: for ( i = 0; i < n; ) {
2197: g = ndvtond(m,nd_ps[i]);
2198: g = nd_separate_head(g,&head);
2199: stat = nd_nf(m,head,g,nd_ps,1,0,&nf);
2200: if ( !stat )
2201: nd_reconstruct(0,0);
2202: else {
2203: if ( DP_Print ) { printf("."); fflush(stdout); }
2204: ndv_free(nd_ps[i]);
2205: nd_removecont(m,nf);
2206: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
2207: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
2208: i++;
2209: }
2210: }
2211: if ( DP_Print ) { printf("\n"); }
2212: for ( a0 = 0, i = 0; i < n; i++ ) {
2213: NEXTNODE(a0,a);
2214: BDY(a) = (pointer)nd_ps[i];
2215: }
2216: NEXT(a) = 0;
2217: return a0;
1.23 noro 2218: }
2219:
1.1 noro 2220: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
2221: {
1.157 noro 2222: ND_pairs d1,nd,cur,head,prev,remove;
1.1 noro 2223:
1.157 noro 2224: if ( !g ) return d;
2225: d = crit_B(d,t);
2226: d1 = nd_newpairs(g,t);
2227: d1 = crit_M(d1);
2228: d1 = crit_F(d1);
2229: if ( do_weyl )
2230: head = d1;
2231: else {
2232: prev = 0; cur = head = d1;
2233: while ( cur ) {
2234: if ( crit_2( cur->i1,cur->i2 ) ) {
2235: remove = cur;
2236: if ( !prev ) head = cur = NEXT(cur);
2237: else cur = NEXT(prev) = NEXT(cur);
2238: FREENDP(remove);
2239: } else {
2240: prev = cur; cur = NEXT(cur);
2241: }
2242: }
2243: }
2244: if ( !d )
2245: return head;
2246: else {
2247: nd = d;
2248: while ( NEXT(nd) ) nd = NEXT(nd);
2249: NEXT(nd) = head;
2250: return d;
2251: }
1.1 noro 2252: }
2253:
1.157 noro 2254:
1.1 noro 2255: ND_pairs nd_newpairs( NODE g, int t )
2256: {
1.157 noro 2257: NODE h;
2258: UINT *dl;
2259: int ts,s;
2260: ND_pairs r,r0;
2261:
2262: dl = DL(nd_psh[t]);
2263: ts = SG(nd_psh[t]) - TD(dl);
2264: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
2265: if ( nd_module && (MPOS(DL(nd_psh[(int)BDY(h)])) != MPOS(dl)) )
2266: continue;
2267: NEXTND_pairs(r0,r);
2268: r->i1 = (int)BDY(h);
2269: r->i2 = t;
2270: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
2271: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
2272: SG(r) = MAX(s,ts) + TD(LCM(r));
2273: }
2274: if ( r0 ) NEXT(r) = 0;
2275: return r0;
1.1 noro 2276: }
2277:
1.157 noro 2278: /* kokokara */
2279:
1.1 noro 2280: ND_pairs crit_B( ND_pairs d, int s )
2281: {
1.157 noro 2282: ND_pairs cur,head,prev,remove;
2283: UINT *t,*tl,*lcm;
2284: int td,tdl;
2285:
2286: if ( !d ) return 0;
2287: t = DL(nd_psh[s]);
2288: prev = 0;
2289: head = cur = d;
2290: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2291: while ( cur ) {
2292: tl = cur->lcm;
2293: if ( ndl_reducible(tl,t)
2294: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
2295: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
2296: remove = cur;
2297: if ( !prev ) {
2298: head = cur = NEXT(cur);
2299: } else {
2300: cur = NEXT(prev) = NEXT(cur);
2301: }
2302: FREENDP(remove);
2303: } else {
2304: prev = cur; cur = NEXT(cur);
2305: }
2306: }
2307: return head;
1.1 noro 2308: }
2309:
2310: ND_pairs crit_M( ND_pairs d1 )
2311: {
1.157 noro 2312: ND_pairs e,d2,d3,dd,p;
2313: UINT *id,*jd;
1.1 noro 2314:
1.157 noro 2315: if ( !d1 ) return d1;
2316: for ( dd = 0, e = d1; e; e = d3 ) {
2317: if ( !(d2 = NEXT(e)) ) {
2318: NEXT(e) = dd;
2319: return e;
2320: }
2321: id = LCM(e);
2322: for ( d3 = 0; d2; d2 = p ) {
2323: p = NEXT(d2);
2324: jd = LCM(d2);
2325: if ( ndl_equal(jd,id) )
2326: ;
2327: else if ( TD(jd) > TD(id) )
2328: if ( ndl_reducible(jd,id) ) continue;
2329: else ;
2330: else if ( ndl_reducible(id,jd) ) goto delit;
2331: NEXT(d2) = d3;
2332: d3 = d2;
2333: }
2334: NEXT(e) = dd;
2335: dd = e;
2336: continue;
2337: /**/
2338: delit: NEXT(d2) = d3;
2339: d3 = d2;
2340: for ( ; p; p = d2 ) {
2341: d2 = NEXT(p);
2342: NEXT(p) = d3;
2343: d3 = p;
2344: }
2345: FREENDP(e);
2346: }
2347: return dd;
1.1 noro 2348: }
2349:
2350: ND_pairs crit_F( ND_pairs d1 )
2351: {
1.157 noro 2352: ND_pairs rest, head,remove;
2353: ND_pairs last, p, r, w;
2354: int s;
2355:
2356: if ( !d1 ) return d1;
2357: for ( head = last = 0, p = d1; NEXT(p); ) {
2358: r = w = equivalent_pairs(p,&rest);
2359: s = SG(r);
2360: w = NEXT(w);
2361: while ( w ) {
2362: if ( crit_2(w->i1,w->i2) ) {
2363: r = w;
2364: w = NEXT(w);
2365: while ( w ) {
2366: remove = w;
2367: w = NEXT(w);
2368: FREENDP(remove);
2369: }
2370: break;
2371: } else if ( SG(w) < s ) {
2372: FREENDP(r);
2373: r = w;
2374: s = SG(r);
2375: w = NEXT(w);
2376: } else {
2377: remove = w;
2378: w = NEXT(w);
2379: FREENDP(remove);
2380: }
2381: }
2382: if ( last ) NEXT(last) = r;
2383: else head = r;
2384: NEXT(last = r) = 0;
2385: p = rest;
2386: if ( !p ) return head;
2387: }
2388: if ( !last ) return p;
2389: NEXT(last) = p;
2390: return head;
1.1 noro 2391: }
2392:
2393: int crit_2( int dp1, int dp2 )
2394: {
1.157 noro 2395: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2396: }
2397:
1.40 noro 2398: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2399: {
1.157 noro 2400: ND_pairs w,p,r,s;
2401: UINT *d;
1.1 noro 2402:
1.157 noro 2403: w = d1;
2404: d = LCM(w);
2405: s = NEXT(w);
2406: NEXT(w) = 0;
2407: for ( r = 0; s; s = p ) {
2408: p = NEXT(s);
2409: if ( ndl_equal(d,LCM(s)) ) {
2410: NEXT(s) = w; w = s;
2411: } else {
2412: NEXT(s) = r; r = s;
2413: }
2414: }
2415: *prest = r;
2416: return w;
1.1 noro 2417: }
2418:
2419: NODE update_base(NODE nd,int ndp)
2420: {
1.157 noro 2421: UINT *dl, *dln;
2422: NODE last, p, head;
1.1 noro 2423:
1.157 noro 2424: dl = DL(nd_psh[ndp]);
2425: for ( head = last = 0, p = nd; p; ) {
2426: dln = DL(nd_psh[(int)BDY(p)]);
2427: if ( ndl_reducible( dln, dl ) ) {
2428: p = NEXT(p);
2429: if ( last ) NEXT(last) = p;
2430: } else {
2431: if ( !last ) head = p;
2432: p = NEXT(last = p);
2433: }
2434: }
2435: head = append_one(head,ndp);
2436: return head;
1.1 noro 2437: }
2438:
2439: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2440: {
1.157 noro 2441: ND_pairs m,ml,p,l;
2442: UINT *lcm;
2443: int s,td,len,tlen,c,c1;
2444:
2445: if ( !(p = NEXT(m = d)) ) {
2446: *prest = p;
2447: NEXT(m) = 0;
2448: return m;
2449: }
2450: s = SG(m);
2451: if ( !NoSugar ) {
2452: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2453: if ( (SG(p) < s)
2454: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2455: ml = l; m = p; s = SG(m);
2456: }
2457: } else {
2458: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2459: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2460: ml = l; m = p; s = SG(m);
2461: }
2462: }
2463: if ( !ml ) *prest = NEXT(m);
2464: else {
2465: NEXT(ml) = NEXT(m);
2466: *prest = d;
2467: }
2468: NEXT(m) = 0;
2469: return m;
1.1 noro 2470: }
2471:
1.63 noro 2472: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2473: {
1.157 noro 2474: int msugar,i;
2475: ND_pairs t,dm0,dm,dr0,dr;
1.63 noro 2476:
1.157 noro 2477: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2478: if ( SG(t) < msugar ) msugar = SG(t);
2479: dm0 = 0; dr0 = 0;
2480: for ( i = 0, t = d; t; t = NEXT(t) )
2481: if ( i < nd_f4_nsp && SG(t) == msugar ) {
2482: if ( dm0 ) NEXT(dm) = t;
2483: else dm0 = t;
2484: dm = t;
2485: i++;
2486: } else {
2487: if ( dr0 ) NEXT(dr) = t;
2488: else dr0 = t;
2489: dr = t;
2490: }
2491: NEXT(dm) = 0;
2492: if ( dr0 ) NEXT(dr) = 0;
2493: *prest = dr0;
2494: return dm0;
1.63 noro 2495: }
2496:
1.77 noro 2497: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2498: {
1.157 noro 2499: int len;
2500: RHist r;
2501: NDV b;
2502:
2503: if ( nd_psn == nd_pslen ) {
2504: nd_pslen *= 2;
2505: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
2506: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
2507: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
2508: nd_bound = (UINT **)
2509: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
2510: }
2511: NEWRHist(r); nd_psh[nd_psn] = r;
2512: nd_ps[nd_psn] = a;
2513: if ( aq ) {
2514: nd_ps_trace[nd_psn] = aq;
2515: register_hcf(aq);
2516: nd_bound[nd_psn] = ndv_compute_bound(aq);
2517: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
2518: } else {
2519: if ( !m ) register_hcf(a);
2520: nd_bound[nd_psn] = ndv_compute_bound(a);
2521: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
2522: }
2523: if ( nd_demand ) {
2524: if ( aq ) {
2525: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2526: nd_ps_trace[nd_psn] = 0;
2527: } else {
2528: ndv_save(nd_ps[nd_psn],nd_psn);
2529: nd_ps[nd_psn] = 0;
2530: }
2531: }
2532: return nd_psn++;
1.1 noro 2533: }
2534:
1.151 noro 2535: void ndv_setup(int mod,int trace,NODE f,int dont_sort,int dont_removecont)
1.1 noro 2536: {
1.157 noro 2537: int i,j,td,len,max;
2538: NODE s,s0,f0;
2539: UINT *d;
2540: RHist r;
2541: NDV *w;
2542: NDV a,am;
2543:
2544: nd_found = 0; nd_notfirst = 0; nd_create = 0;
2545:
2546: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2547: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2548: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2549: if ( !dont_sort ) {
2550: /* XXX heuristic */
2551: if ( !nd_ord->id && (nd_ord->ord.simple<2) )
2552: qsort(w,nd_psn,sizeof(NDV),
2553: (int (*)(const void *,const void *))ndv_compare_rev);
2554: else
2555: qsort(w,nd_psn,sizeof(NDV),
2556: (int (*)(const void *,const void *))ndv_compare);
2557: }
2558: nd_pslen = 2*nd_psn;
2559: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
2560: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
2561: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
2562: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
2563: nd_hcf = 0;
2564:
2565: if ( trace && nd_vc )
2566: makesubst(nd_vc,&nd_subst);
2567: else
2568: nd_subst = 0;
2569:
2570: if ( !nd_red )
2571: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
2572: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
2573: for ( i = 0; i < nd_psn; i++ ) {
2574: if ( trace ) {
2575: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
2576: if ( !dont_removecont) ndv_removecont(0,a);
2577: register_hcf(a);
2578: am = nd_ps[i] = ndv_dup(mod,a);
2579: ndv_mod(mod,am);
2580: ndv_removecont(mod,am);
2581: } else {
2582: a = nd_ps[i] = ndv_dup(mod,w[i]);
2583: if ( mod || !dont_removecont ) ndv_removecont(mod,a);
2584: if ( !mod ) register_hcf(a);
2585: }
2586: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
2587: nd_bound[i] = ndv_compute_bound(a);
2588: nd_psh[i] = r;
2589: if ( nd_demand ) {
2590: if ( trace ) {
2591: ndv_save(nd_ps_trace[i],i);
2592: nd_ps_trace[i] = 0;
2593: } else {
2594: ndv_save(nd_ps[i],i);
2595: nd_ps[i] = 0;
2596: }
2597: }
2598: }
1.20 noro 2599: }
2600:
1.119 noro 2601: struct order_spec *append_block(struct order_spec *spec,
2602: int nv,int nalg,int ord);
2603:
1.121 noro 2604: extern VECT current_dl_weight_vector_obj;
2605: static VECT prev_weight_vector_obj;
2606:
1.120 noro 2607: void preprocess_algcoef(VL vv,VL av,struct order_spec *ord,LIST f,
1.157 noro 2608: struct order_spec **ord1p,LIST *f1p,NODE *alistp)
1.120 noro 2609: {
1.157 noro 2610: NODE alist,t,s,r0,r,arg;
2611: VL tv;
2612: P poly;
2613: DP d;
2614: Alg alpha,dp;
2615: DAlg inv,da,hc;
2616: MP m;
2617: int i,nvar,nalg,n;
2618: NumberField nf;
2619: LIST f1,f2;
2620: struct order_spec *current_spec;
2621: VECT obj,obj0;
2622: Obj tmp;
2623:
2624: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++);
2625: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++);
2626:
2627: for ( alist = 0, tv = av; tv; tv = NEXT(tv) ) {
2628: NEXTNODE(alist,t); MKV(tv->v,poly);
2629: MKAlg(poly,alpha); BDY(t) = (pointer)alpha;
2630: tv->v = tv->v->priv;
2631: }
2632: NEXT(t) = 0;
2633:
2634: /* simplification, makeing polynomials monic */
2635: setfield_dalg(alist);
2636: obj_algtodalg(f,&f1);
2637: for ( t = BDY(f); t; t = NEXT(t) ) {
2638: initd(ord); ptod(vv,vv,(P)BDY(t),&d);
2639: hc = (DAlg)BDY(d)->c;
2640: if ( NID(hc) == N_DA ) {
2641: invdalg(hc,&inv);
2642: for ( m = BDY(d); m; m = NEXT(m) ) {
2643: muldalg(inv,(DAlg)m->c,&da); m->c = (P)da;
2644: }
2645: }
2646: initd(ord); dtop(vv,vv,d,&poly); BDY(f) = (pointer)poly;
2647: }
2648: obj_dalgtoalg(f1,&f);
2649:
2650: /* append alg vars to the var list */
2651: for ( tv = vv; NEXT(tv); tv = NEXT(tv) );
2652: NEXT(tv) = av;
2653:
2654: /* append a block to ord */
2655: *ord1p = append_block(ord,nvar,nalg,2);
2656:
2657: /* create generator list */
2658: nf = get_numberfield();
2659: for ( i = nalg-1, t = BDY(f); i >= 0; i-- ) {
2660: MKAlg(nf->defpoly[i],dp);
2661: MKNODE(s,dp,t); t = s;
2662: }
2663: MKLIST(f1,t);
2664: *alistp = alist;
2665: algobjtorat(f1,f1p);
2666:
2667: /* creating a new weight vector */
2668: prev_weight_vector_obj = obj0 = current_dl_weight_vector_obj;
2669: n = nvar+nalg+1;
2670: MKVECT(obj,n);
2671: if ( obj0 && obj0->len == nvar )
2672: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = BDY(obj0)[i];
2673: else
2674: for ( i = 0; i < nvar; i++ ) BDY(obj)[i] = (pointer)ONE;
2675: for ( i = 0; i < nalg; i++ ) BDY(obj)[i+nvar] = 0;
2676: BDY(obj)[n-1] = (pointer)ONE;
2677: arg = mknode(1,obj);
2678: Pdp_set_weight(arg,&tmp);
1.121 noro 2679: }
2680:
2681: NODE postprocess_algcoef(VL av,NODE alist,NODE r)
2682: {
1.157 noro 2683: NODE s,t,u0,u;
2684: P p;
2685: VL tv;
2686: Obj obj,tmp;
2687: NODE arg;
2688:
2689: u0 = 0;
2690: for ( t = r; t; t = NEXT(t) ) {
2691: p = (P)BDY(t);
2692: for ( tv = av, s = alist; tv; tv = NEXT(tv), s = NEXT(s) ) {
2693: substr(CO,0,(Obj)p,tv->v,(Obj)BDY(s),&obj); p = (P)obj;
2694: }
2695: if ( OID(p) == O_P || (OID(p) == O_N && NID((Num)p) != N_A) ) {
2696: NEXTNODE(u0,u);
2697: BDY(u) = (pointer)p;
2698: }
2699: }
2700: arg = mknode(1,prev_weight_vector_obj);
2701: Pdp_set_weight(arg,&tmp);
1.121 noro 2702:
1.157 noro 2703: return u0;
1.120 noro 2704: }
2705:
1.63 noro 2706: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2707: {
1.157 noro 2708: VL tv,fv,vv,vc,av;
2709: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2710: int e,max,nvar,i;
2711: NDV b;
2712: int ishomo,nalg,mrank,trank;
2713: Alg alpha,dp;
2714: P p,zp;
2715: Q dmy;
2716: LIST f1,f2,zpl;
2717: Obj obj;
2718: NumberField nf;
2719: struct order_spec *ord1;
1.1 noro 2720:
1.157 noro 2721: if ( !m && Demand ) nd_demand = 1;
2722: else nd_demand = 0;
1.78 noro 2723:
1.157 noro 2724: if ( DP_Multiple )
2725: nd_scale = ((double)DP_Multiple)/(double)(Denominator?Denominator:1);
1.103 noro 2726: #if 0
1.157 noro 2727: ndv_alloc = 0;
1.103 noro 2728: #endif
1.157 noro 2729: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
2730: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2731: switch ( ord->id ) {
2732: case 1:
2733: if ( ord->nv != nvar )
2734: error("nd_{gr,f4} : invalid order specification");
2735: break;
2736: default:
2737: break;
2738: }
2739: nd_nalg = 0;
2740: av = 0;
2741: if ( !m ) {
2742: get_algtree((Obj)f,&av);
2743: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2744: nd_ntrans = nvar;
2745: nd_nalg = nalg;
2746: /* #i -> t#i */
2747: if ( nalg ) {
2748: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2749: ord = ord1;
2750: f = f1;
2751: }
2752: nvar += nalg;
2753: }
2754: nd_init_ord(ord);
2755: mrank = 0;
2756: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2757: for ( tv = vv; tv; tv = NEXT(tv) ) {
2758: if ( nd_module ) {
2759: s = BDY((LIST)BDY(t));
2760: trank = length(s);
2761: mrank = MAX(mrank,trank);
2762: for ( ; s; s = NEXT(s) ) {
2763: e = getdeg(tv->v,(P)BDY(s));
2764: max = MAX(e,max);
2765: }
2766: } else {
2767: e = getdeg(tv->v,(P)BDY(t));
2768: max = MAX(e,max);
2769: }
2770: }
2771: nd_setup_parameters(nvar,max);
2772: ishomo = 1;
2773: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2774: if ( nd_module ) {
2775: pltozpl((LIST)BDY(t),&dmy,&zpl);
2776: b = (pointer)pltondv(CO,vv,zpl);
2777: } else {
2778: ptozp((P)BDY(t),1,&dmy,&zp);
2779: b = (pointer)ptondv(CO,vv,zp);
2780: }
2781: if ( ishomo )
2782: ishomo = ishomo && ndv_ishomo(b);
2783: if ( m ) ndv_mod(m,b);
2784: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2785: }
2786: if ( fd0 ) NEXT(fd) = 0;
2787: ndv_setup(m,0,fd0,0,0);
2788: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
2789: nd_demand = 0;
2790: x = ndv_reducebase(x);
2791: x = ndv_reduceall(m,x);
2792: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2793: NEXTNODE(r0,r);
2794: if ( nd_module ) BDY(r) = ndvtopl(m,CO,vv,BDY(t),mrank);
2795: else BDY(r) = ndvtop(m,CO,vv,BDY(t));
2796: }
2797: if ( r0 ) NEXT(r) = 0;
2798: if ( nalg )
2799: r0 = postprocess_algcoef(av,alist,r0);
2800: MKLIST(*rp,r0);
1.103 noro 2801: #if 0
1.157 noro 2802: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 2803: #endif
1.127 noro 2804: }
2805:
2806: void nd_gr_postproc(LIST f,LIST v,int m,struct order_spec *ord,int do_check,LIST *rp)
2807: {
1.157 noro 2808: VL tv,fv,vv,vc,av;
2809: NODE fd,fd0,r,r0,t,x,s,xx,alist;
2810: int e,max,nvar,i;
2811: NDV b;
2812: int ishomo,nalg;
2813: Alg alpha,dp;
2814: P p,zp;
2815: Q dmy;
2816: LIST f1,f2;
2817: Obj obj;
2818: NumberField nf;
2819: struct order_spec *ord1;
2820:
2821: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
2822: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2823: switch ( ord->id ) {
2824: case 1:
2825: if ( ord->nv != nvar )
2826: error("nd_check : invalid order specification");
2827: break;
2828: default:
2829: break;
2830: }
2831: nd_nalg = 0;
2832: av = 0;
2833: if ( !m ) {
2834: get_algtree((Obj)f,&av);
2835: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2836: nd_ntrans = nvar;
2837: nd_nalg = nalg;
2838: /* #i -> t#i */
2839: if ( nalg ) {
2840: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2841: ord = ord1;
2842: f = f1;
2843: }
2844: nvar += nalg;
2845: }
2846: nd_init_ord(ord);
2847: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2848: for ( tv = vv; tv; tv = NEXT(tv) ) {
2849: e = getdeg(tv->v,(P)BDY(t));
2850: max = MAX(e,max);
2851: }
2852: nd_setup_parameters(nvar,max);
2853: ishomo = 1;
2854: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2855: ptozp((P)BDY(t),1,&dmy,&zp);
2856: b = (pointer)ptondv(CO,vv,zp);
2857: if ( ishomo )
2858: ishomo = ishomo && ndv_ishomo(b);
2859: if ( m ) ndv_mod(m,b);
2860: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
2861: }
2862: if ( fd0 ) NEXT(fd) = 0;
2863: ndv_setup(m,0,fd0,0,1);
2864: for ( x = 0, i = 0; i < nd_psn; i++ )
2865: x = update_base(x,i);
2866: if ( do_check ) {
2867: x = nd_gb(m,ishomo,1);
2868: if ( !x ) {
2869: *rp = 0;
2870: return;
2871: }
2872: } else {
2873: for ( t = x; t; t = NEXT(t) )
2874: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
2875: }
2876: x = ndv_reducebase(x);
2877: x = ndv_reduceall(m,x);
2878: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
2879: NEXTNODE(r0,r);
2880: BDY(r) = ndvtop(m,CO,vv,BDY(t));
2881: }
2882: if ( r0 ) NEXT(r) = 0;
2883: if ( nalg )
2884: r0 = postprocess_algcoef(av,alist,r0);
2885: MKLIST(*rp,r0);
1.20 noro 2886: }
2887:
1.133 noro 2888: void nd_gr_trace(LIST f,LIST v,int trace,int homo,int f4,struct order_spec *ord,LIST *rp)
1.20 noro 2889: {
1.157 noro 2890: VL tv,fv,vv,vc,av;
2891: NODE fd,fd0,in0,in,r,r0,t,s,cand,alist;
2892: int m,nocheck,nvar,mindex,e,max;
2893: NDV c;
2894: NMV a;
2895: P p,zp;
2896: Q dmy;
2897: EPOS oepos;
1.158 ! noro 2898: int obpe,oadv,wmax,i,len,cbpe,ishomo,nalg,mrank,trank;
1.157 noro 2899: Alg alpha,dp;
2900: P poly;
1.158 ! noro 2901: LIST f1,f2,zpl;
1.157 noro 2902: Obj obj;
2903: NumberField nf;
2904: struct order_spec *ord1;
2905: struct oEGT eg_check,eg0,eg1;
2906:
2907: if ( DP_Multiple )
2908: nd_scale = ((double)DP_Multiple)/(double)(Denominator?Denominator:1);
2909:
2910: get_vars((Obj)f,&fv); pltovl(v,&vv); vlminus(fv,vv,&nd_vc);
2911: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
2912: switch ( ord->id ) {
2913: case 1:
2914: if ( ord->nv != nvar )
2915: error("nd_gr_trace : invalid order specification");
2916: break;
2917: default:
2918: break;
2919: }
2920:
2921: get_algtree((Obj)f,&av);
2922: for ( nalg = 0, tv = av; tv; tv = NEXT(tv), nalg++ );
2923: nd_ntrans = nvar;
2924: nd_nalg = nalg;
2925: /* #i -> t#i */
2926: if ( nalg ) {
2927: preprocess_algcoef(vv,av,ord,f,&ord1,&f1,&alist);
2928: ord = ord1;
2929: f = f1;
2930: }
2931: nvar += nalg;
2932:
2933: nocheck = 0;
2934: mindex = 0;
2935:
2936: if ( Demand ) nd_demand = 1;
2937: else nd_demand = 0;
2938:
2939: /* setup modulus */
2940: if ( trace < 0 ) {
2941: trace = -trace;
2942: nocheck = 1;
2943: }
2944: m = trace > 1 ? trace : get_lprime(mindex);
1.158 ! noro 2945: nd_init_ord(ord);
! 2946: mrank = 0;
1.157 noro 2947: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2948: for ( tv = vv; tv; tv = NEXT(tv) ) {
1.158 ! noro 2949: if ( nd_module ) {
! 2950: s = BDY((LIST)BDY(t));
! 2951: trank = length(s);
! 2952: mrank = MAX(mrank,trank);
! 2953: for ( ; s; s = NEXT(s) ) {
! 2954: e = getdeg(tv->v,(P)BDY(s));
! 2955: max = MAX(e,max);
! 2956: }
! 2957: } else {
! 2958: e = getdeg(tv->v,(P)BDY(t));
! 2959: max = MAX(e,max);
! 2960: }
1.157 noro 2961: }
2962: nd_setup_parameters(nvar,max);
2963: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2964: ishomo = 1;
2965: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.158 ! noro 2966: if ( nd_module ) {
! 2967: pltozpl((LIST)BDY(t),&dmy,&zpl);
! 2968: c = (pointer)pltondv(CO,vv,zpl);
! 2969: } else {
! 2970: ptozp((P)BDY(t),1,&dmy,&zp);
! 2971: c = (pointer)ptondv(CO,vv,zp);
! 2972: }
1.157 noro 2973: if ( ishomo )
2974: ishomo = ishomo && ndv_ishomo(c);
2975: if ( c ) {
2976: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2977: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
2978: }
2979: }
2980: if ( in0 ) NEXT(in) = 0;
2981: if ( fd0 ) NEXT(fd) = 0;
2982: if ( !ishomo && homo ) {
2983: for ( t = in0, wmax = max; t; t = NEXT(t) ) {
2984: c = (NDV)BDY(t); len = LEN(c);
2985: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2986: wmax = MAX(TD(DL(a)),wmax);
2987: }
2988: homogenize_order(ord,nvar,&ord1);
2989: nd_init_ord(ord1);
2990: nd_setup_parameters(nvar+1,wmax);
2991: for ( t = fd0; t; t = NEXT(t) )
2992: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
2993: }
2994: while ( 1 ) {
2995: if ( Demand )
2996: nd_demand = 1;
2997: ndv_setup(m,1,fd0,0,0);
2998: cand = f4?nd_f4_trace(m):nd_gb_trace(m,ishomo || homo);
2999: if ( !cand ) {
3000: /* failure */
3001: if ( trace > 1 ) { *rp = 0; return; }
3002: else m = get_lprime(++mindex);
3003: continue;
3004: }
3005: if ( !ishomo && homo ) {
3006: /* dehomogenization */
3007: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
3008: nd_init_ord(ord);
3009: nd_setup_parameters(nvar,0);
3010: }
3011: nd_demand = 0;
3012: cand = ndv_reducebase(cand);
3013: cand = ndv_reduceall(0,cand);
3014: cbpe = nd_bpe;
3015: if ( nocheck )
3016: break;
3017: get_eg(&eg0);
3018: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
3019: /* success */
3020: break;
3021: else if ( trace > 1 ) {
3022: /* failure */
3023: *rp = 0; return;
3024: } else {
3025: /* try the next modulus */
3026: m = get_lprime(++mindex);
3027: /* reset the parameters */
3028: if ( !ishomo && homo ) {
3029: nd_init_ord(ord1);
3030: nd_setup_parameters(nvar+1,wmax);
3031: } else {
3032: nd_init_ord(ord);
3033: nd_setup_parameters(nvar,max);
3034: }
3035: }
3036: }
3037: get_eg(&eg1); init_eg(&eg_check); add_eg(&eg_check,&eg0,&eg1);
3038: if ( DP_Print )
3039: fprintf(asir_out,"check=%fsec\n",eg_check.exectime+eg_check.gctime);
3040: /* dp->p */
3041: nd_bpe = cbpe;
3042: nd_setup_parameters(nd_nvar,0);
1.158 ! noro 3043: for ( r = cand; r; r = NEXT(r) ) {
! 3044: if ( nd_module ) BDY(r) = ndvtopl(0,CO,vv,BDY(r),mrank);
! 3045: else BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
! 3046: }
1.157 noro 3047: if ( nalg )
3048: cand = postprocess_algcoef(av,alist,cand);
3049: MKLIST(*rp,cand);
3050: }
1.52 noro 3051:
1.157 noro 3052: /* XXX : module element is not considered */
1.1 noro 3053:
1.61 noro 3054: void dltondl(int n,DL dl,UINT *r)
1.1 noro 3055: {
1.157 noro 3056: UINT *d;
3057: int i,j,l,s,ord_l;
3058: struct order_pair *op;
3059:
3060: d = dl->d;
3061: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
3062: if ( nd_blockmask ) {
3063: l = nd_blockmask->n;
3064: op = nd_blockmask->order_pair;
3065: for ( j = 0, s = 0; j < l; j++ ) {
3066: ord_l = op[j].length;
3067: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
3068: }
3069: TD(r) = ndl_weight(r);
3070: ndl_weight_mask(r);
3071: } else {
3072: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
3073: TD(r) = ndl_weight(r);
3074: }
1.1 noro 3075: }
3076:
1.61 noro 3077: DL ndltodl(int n,UINT *ndl)
1.1 noro 3078: {
1.157 noro 3079: DL dl;
3080: int *d;
3081: int i,j,l,s,ord_l;
3082: struct order_pair *op;
3083:
3084: NEWDL(dl,n);
3085: dl->td = TD(ndl);
3086: d = dl->d;
3087: if ( nd_blockmask ) {
3088: l = nd_blockmask->n;
3089: op = nd_blockmask->order_pair;
3090: for ( j = 0, s = 0; j < l; j++ ) {
3091: ord_l = op[j].length;
3092: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
3093: }
3094: } else {
3095: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
3096: }
3097: return dl;
1.1 noro 3098: }
3099:
1.61 noro 3100: void ndl_print(UINT *dl)
1.1 noro 3101: {
1.157 noro 3102: int n;
3103: int i,j,l,ord_l,s,s0;
3104: struct order_pair *op;
3105:
3106: n = nd_nvar;
3107: printf("<<");
3108: if ( nd_blockmask ) {
3109: l = nd_blockmask->n;
3110: op = nd_blockmask->order_pair;
3111: for ( j = 0, s = s0 = 0; j < l; j++ ) {
3112: ord_l = op[j].length;
3113: for ( i = 0; i < ord_l; i++, s++ )
3114: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
3115: }
3116: } else {
3117: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
3118: }
3119: printf(">>");
3120: if ( MPOS(dl) )
3121: printf("*e%d",MPOS(dl));
1.1 noro 3122: }
3123:
3124: void nd_print(ND p)
3125: {
1.157 noro 3126: NM m;
1.1 noro 3127:
1.157 noro 3128: if ( !p )
3129: printf("0\n");
3130: else {
3131: for ( m = BDY(p); m; m = NEXT(m) ) {
3132: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3133: else printf("+%d*",CM(m));
3134: ndl_print(DL(m));
3135: }
3136: printf("\n");
3137: }
1.1 noro 3138: }
3139:
1.113 noro 3140: void nd_print_q(ND p)
1.16 noro 3141: {
1.157 noro 3142: NM m;
1.16 noro 3143:
1.157 noro 3144: if ( !p )
3145: printf("0\n");
3146: else {
3147: for ( m = BDY(p); m; m = NEXT(m) ) {
3148: printf("+");
3149: printexpr(CO,(Obj)CQ(m));
3150: printf("*");
3151: ndl_print(DL(m));
3152: }
3153: printf("\n");
3154: }
1.16 noro 3155: }
3156:
1.1 noro 3157: void ndp_print(ND_pairs d)
3158: {
1.157 noro 3159: ND_pairs t;
1.1 noro 3160:
1.157 noro 3161: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
3162: printf("\n");
1.1 noro 3163: }
3164:
1.20 noro 3165: void nd_removecont(int mod,ND p)
1.16 noro 3166: {
1.157 noro 3167: int i,n;
3168: Q *w;
3169: Q dvr,t;
3170: NM m;
3171: struct oVECT v;
3172: N q,r;
3173:
3174: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
3175: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
3176: else {
3177: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
3178: w = (Q *)ALLOCA(n*sizeof(Q));
3179: v.len = n;
3180: v.body = (pointer *)w;
3181: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
3182: removecont_array((P *)w,n,1);
3183: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
3184: }
1.16 noro 3185: }
3186:
1.21 noro 3187: void nd_removecont2(ND p1,ND p2)
3188: {
1.157 noro 3189: int i,n1,n2,n;
3190: Q *w;
3191: Q dvr,t;
3192: NM m;
3193: struct oVECT v;
3194: N q,r;
3195:
3196: n1 = nd_length(p1);
3197: n2 = nd_length(p2);
3198: n = n1+n2;
3199: w = (Q *)ALLOCA(n*sizeof(Q));
3200: v.len = n;
3201: v.body = (pointer *)w;
3202: i = 0;
3203: if ( p1 )
3204: for ( m = BDY(p1); i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
3205: if ( p2 )
3206: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
3207: removecont_array((P *)w,n,1);
3208: i = 0;
3209: if ( p1 )
3210: for ( m = BDY(p1); i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
3211: if ( p2 )
3212: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.21 noro 3213: }
3214:
1.20 noro 3215: void ndv_removecont(int mod,NDV p)
1.16 noro 3216: {
1.157 noro 3217: int i,len,all_p;
3218: Q *c;
3219: P *w;
3220: Q dvr,t;
3221: P g,cont,tp;
3222: NMV m;
3223:
3224: if ( mod == -1 )
3225: ndv_mul_c(mod,p,_invsf(HCM(p)));
3226: else if ( mod )
3227: ndv_mul_c(mod,p,invm(HCM(p),mod));
3228: else {
3229: len = p->len;
3230: w = (P *)ALLOCA(len*sizeof(P));
3231: c = (Q *)ALLOCA(len*sizeof(Q));
3232: for ( m = BDY(p), all_p = 1, i = 0; i < len; NMV_ADV(m), i++ ) {
3233: ptozp(CP(m),1,&c[i],&w[i]);
3234: all_p = all_p && !NUM(w[i]);
3235: }
3236: if ( all_p ) {
3237: qltozl(c,len,&dvr); nd_heu_nezgcdnpz(nd_vc,w,len,1,&g);
3238: mulp(nd_vc,(P)dvr,g,&cont);
3239: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
3240: divsp(nd_vc,CP(m),cont,&tp); CP(m) = tp;
3241: }
3242: } else {
3243: sortbynm((Q *)c,len);
3244: qltozl((Q *)c,len,&dvr);
3245: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
3246: divsp(nd_vc,CP(m),(P)dvr,&tp); CP(m) = tp;
3247: }
3248: }
3249: }
1.21 noro 3250: }
3251:
1.157 noro 3252: /* koko */
3253:
1.61 noro 3254: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
3255: {
1.157 noro 3256: int len,i,max;
3257: NMV m,mr0,mr,t;
1.61 noro 3258:
1.157 noro 3259: len = p->len;
3260: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
3261: max = MAX(max,TD(DL(m)));
3262: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3263: m = (NMV)((char *)mr0+(len-1)*oadv);
3264: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3265: t = (NMV)ALLOCA(nmv_adv);
3266: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3267: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
3268: CQ(mr) = CQ(m);
3269: ndl_copy(DL(t),DL(mr));
3270: }
3271: NV(p)++;
3272: BDY(p) = mr0;
1.61 noro 3273: }
3274:
1.45 noro 3275: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 3276: {
1.157 noro 3277: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
3278: Q *w;
3279: Q dvr,t;
3280: NMV m,r;
3281:
3282: len = p->len;
3283: newnvar = nd_nvar-1;
3284: newexporigin = nd_get_exporigin(ord);
3285: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
3286: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
3287: ndl_dehomogenize(DL(m));
3288: if ( newwpd != nd_wpd ) {
3289: newadv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT));
3290: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
3291: CQ(r) = CQ(m);
3292: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
3293: adj = nd_exporigin-newexporigin;
3294: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
3295: }
3296: }
3297: NV(p)--;
1.23 noro 3298: }
3299:
1.150 noro 3300: void nd_heu_nezgcdnpz(VL vl,P *pl,int m,int full,P *pr)
3301: {
1.157 noro 3302: int i;
3303: P *tpl,*tpl1;
3304: NODE l;
3305: P h,gcd,t;
3306:
3307: tpl = (P *)ALLOCA(m*sizeof(P));
3308: tpl1 = (P *)ALLOCA(m*sizeof(P));
3309: bcopy(pl,tpl,m*sizeof(P));
3310: gcd = (P)ONE;
3311: for ( l = nd_hcf; l; l = NEXT(l) ) {
3312: h = (P)BDY(l);
3313: while ( 1 ) {
3314: for ( i = 0; i < m; i++ )
3315: if ( !divtpz(vl,tpl[i],h,&tpl1[i]) )
3316: break;
3317: if ( i == m ) {
3318: bcopy(tpl1,tpl,m*sizeof(P));
3319: mulp(vl,gcd,h,&t); gcd = t;
3320: } else
3321: break;
3322: }
3323: }
3324: if ( DP_Print > 2 ){fprintf(asir_out,"[%d]",nmonop(gcd)); fflush(asir_out);}
3325: if ( full ) {
3326: heu_nezgcdnpz(vl,tpl,m,&t);
3327: mulp(vl,gcd,t,pr);
3328: } else
3329: *pr = gcd;
1.150 noro 3330: }
3331:
3332: void removecont_array(P *p,int n,int full)
1.146 noro 3333: {
1.157 noro 3334: int all_p,all_q,i;
3335: Q *c;
3336: P *w;
3337: P t,s;
3338:
3339: for ( all_q = 1, i = 0; i < n; i++ )
3340: all_q = all_q && NUM(p[i]);
3341: if ( all_q ) {
3342: removecont_array_q((Q *)p,n);
3343: } else {
3344: c = (Q *)ALLOCA(n*sizeof(Q));
3345: w = (P *)ALLOCA(n*sizeof(P));
3346: for ( i = 0; i < n; i++ ) {
3347: ptozp(p[i],1,&c[i],&w[i]);
3348: }
3349: removecont_array_q(c,n);
3350: nd_heu_nezgcdnpz(nd_vc,w,n,full,&t);
3351: for ( i = 0; i < n; i++ ) {
3352: divsp(nd_vc,w[i],t,&s); mulp(nd_vc,s,(P)c[i],&p[i]);
3353: }
3354: }
1.146 noro 3355: }
3356:
3357: void removecont_array_q(Q *c,int n)
1.21 noro 3358: {
1.157 noro 3359: struct oVECT v;
3360: Q d0,d1,a,u,u1,gcd;
3361: int i,j;
3362: N qn,rn,gn;
3363: Q *q,*r;
3364:
3365: q = (Q *)ALLOCA(n*sizeof(Q));
3366: r = (Q *)ALLOCA(n*sizeof(Q));
3367: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
3368: igcdv_estimate(&v,&d0);
3369: for ( i = 0; i < n; i++ ) {
3370: divn(NM(c[i]),NM(d0),&qn,&rn);
3371: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
3372: NTOQ(rn,SGN(c[i]),r[i]);
3373: }
3374: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
3375: if ( i < n ) {
3376: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
3377: igcdv(&v,&d1);
3378: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
3379: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
3380: for ( i = 0; i < n; i++ ) {
3381: mulq(a,q[i],&u);
3382: if ( r[i] ) {
3383: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
3384: addq(u,u1,&q[i]);
3385: } else
3386: q[i] = u;
3387: }
3388: }
3389: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 3390: }
3391:
1.19 noro 3392: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 3393: {
1.157 noro 3394: NM m;
3395: int c,c1;
1.1 noro 3396:
1.157 noro 3397: if ( !p ) return;
3398: if ( mul == 1 ) return;
3399: if ( mod == -1 )
3400: for ( m = BDY(p); m; m = NEXT(m) )
3401: CM(m) = _mulsf(CM(m),mul);
3402: else
3403: for ( m = BDY(p); m; m = NEXT(m) ) {
3404: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3405: }
1.1 noro 3406: }
3407:
1.146 noro 3408: void nd_mul_c_q(ND p,P mul)
1.16 noro 3409: {
1.157 noro 3410: NM m;
3411: P c;
1.16 noro 3412:
1.157 noro 3413: if ( !p ) return;
3414: if ( UNIQ(mul) ) return;
3415: for ( m = BDY(p); m; m = NEXT(m) ) {
3416: mulp(nd_vc,CP(m),mul,&c); CP(m) = c;
3417: }
1.16 noro 3418: }
3419:
1.61 noro 3420: void nd_mul_c_p(VL vl,ND p,P mul)
3421: {
1.157 noro 3422: NM m;
3423: P c;
1.61 noro 3424:
1.157 noro 3425: if ( !p ) return;
3426: for ( m = BDY(p); m; m = NEXT(m) ) {
3427: mulp(vl,CP(m),mul,&c); CP(m) = c;
3428: }
1.61 noro 3429: }
3430:
1.1 noro 3431: void nd_free(ND p)
3432: {
1.157 noro 3433: NM t,s;
1.1 noro 3434:
1.157 noro 3435: if ( !p ) return;
3436: t = BDY(p);
3437: while ( t ) {
3438: s = NEXT(t);
3439: FREENM(t);
3440: t = s;
3441: }
3442: FREEND(p);
1.1 noro 3443: }
3444:
1.23 noro 3445: void ndv_free(NDV p)
3446: {
1.157 noro 3447: GC_free(BDY(p));
1.23 noro 3448: }
3449:
1.61 noro 3450: void nd_append_red(UINT *d,int i)
1.1 noro 3451: {
1.157 noro 3452: RHist m,m0;
3453: int h;
1.1 noro 3454:
1.157 noro 3455: NEWRHist(m);
3456: h = ndl_hash_value(d);
3457: m->index = i;
3458: ndl_copy(d,DL(m));
3459: NEXT(m) = nd_red[h];
3460: nd_red[h] = m;
1.1 noro 3461: }
3462:
1.61 noro 3463: UINT *ndv_compute_bound(NDV p)
1.1 noro 3464: {
1.157 noro 3465: UINT *d1,*d2,*t;
3466: UINT u;
3467: int i,j,k,l,len,ind;
3468: NMV m;
3469:
3470: if ( !p )
3471: return 0;
3472: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3473: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3474: len = LEN(p);
3475: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
3476: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
3477: ndl_lcm_nocheck(DL(m),d1,d2);
3478: t = d1; d1 = d2; d2 = t;
3479: }
3480: l = nd_nvar+31;
3481: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3482: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3483: u = d1[i];
3484: k = (nd_epw-1)*nd_bpe;
3485: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3486: t[ind] = (u>>k)&nd_mask0;
3487: }
3488: for ( ; ind < l; ind++ ) t[ind] = 0;
3489: return t;
1.1 noro 3490: }
3491:
1.99 noro 3492: UINT *nd_compute_bound(ND p)
3493: {
1.157 noro 3494: UINT *d1,*d2,*t;
3495: UINT u;
3496: int i,j,k,l,len,ind;
3497: NM m;
3498:
3499: if ( !p )
3500: return 0;
3501: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3502: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3503: len = LEN(p);
3504: m = BDY(p); ndl_copy(DL(m),d1); m = NEXT(m);
3505: for ( m = NEXT(m); m; m = NEXT(m) ) {
3506: ndl_lcm(DL(m),d1,d2);
3507: t = d1; d1 = d2; d2 = t;
3508: }
3509: l = nd_nvar+31;
3510: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
3511: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
3512: u = d1[i];
3513: k = (nd_epw-1)*nd_bpe;
3514: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
3515: t[ind] = (u>>k)&nd_mask0;
3516: }
3517: for ( ; ind < l; ind++ ) t[ind] = 0;
3518: return t;
1.99 noro 3519: }
3520:
1.157 noro 3521: /* if nd_module == 1 then d[nd_exporigin-1] indicates the position */
3522: /* of a term. In this case we need additional 1 word. */
3523:
1.48 noro 3524: int nd_get_exporigin(struct order_spec *ord)
3525: {
1.157 noro 3526: switch ( ord->id ) {
3527: case 0: case 2: case 256: case 258:
3528: return 1+nd_module;
3529: case 1: case 257:
3530: /* block order */
3531: /* poly ring d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
3532: /* module d[0]:weight d[1]:w0,...,d[nd_exporigin-2]:w(n-1) */
3533: return ord->ord.block.length+1+nd_module;
3534: case 3: case 259:
3535: error("nd_get_exporigin : composite order is not supported yet.");
3536: }
1.48 noro 3537: }
3538:
1.61 noro 3539: void nd_setup_parameters(int nvar,int max) {
1.157 noro 3540: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
3541: struct order_pair *op;
1.48 noro 3542:
1.157 noro 3543: nd_nvar = nvar;
3544: if ( max ) {
3545: /* XXX */
3546: if ( do_weyl ) nd_bpe = 32;
3547: else if ( max < 2 ) nd_bpe = 1;
3548: else if ( max < 4 ) nd_bpe = 2;
3549: else if ( max < 8 ) nd_bpe = 3;
3550: else if ( max < 16 ) nd_bpe = 4;
3551: else if ( max < 32 ) nd_bpe = 5;
3552: else if ( max < 64 ) nd_bpe = 6;
3553: else if ( max < 256 ) nd_bpe = 8;
3554: else if ( max < 1024 ) nd_bpe = 10;
3555: else if ( max < 65536 ) nd_bpe = 16;
3556: else nd_bpe = 32;
3557: }
3558: nd_epw = (sizeof(UINT)*8)/nd_bpe;
3559: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
3560: nd_exporigin = nd_get_exporigin(nd_ord);
3561: wpd = nd_exporigin+elen;
3562: if ( nd_module )
3563: nd_mpos = nd_exporigin-1;
3564: else
3565: nd_mpos = -1;
3566: if ( wpd != nd_wpd ) {
3567: nd_free_private_storage();
3568: nd_wpd = wpd;
3569: }
3570: if ( nd_bpe < 32 ) {
3571: nd_mask0 = (1<<nd_bpe)-1;
3572: } else {
3573: nd_mask0 = 0xffffffff;
3574: }
3575: bzero(nd_mask,sizeof(nd_mask));
3576: nd_mask1 = 0;
3577: for ( i = 0; i < nd_epw; i++ ) {
3578: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
3579: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
3580: }
3581: nmv_adv = ROUND_FOR_ALIGN(sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT));
3582: nd_epos = nd_create_epos(nd_ord);
3583: nd_blockmask = nd_create_blockmask(nd_ord);
3584: nd_work_vector = (int *)REALLOC(nd_work_vector,nd_nvar*sizeof(int));
1.1 noro 3585: }
3586:
1.103 noro 3587: ND_pairs nd_reconstruct(int trace,ND_pairs d)
1.1 noro 3588: {
1.157 noro 3589: int i,obpe,oadv,h;
3590: static NM prev_nm_free_list;
3591: static ND_pairs prev_ndp_free_list;
3592: RHist mr0,mr;
3593: RHist r;
3594: RHist *old_red;
3595: ND_pairs s0,s,t;
3596: EPOS oepos;
3597:
3598: obpe = nd_bpe;
3599: oadv = nmv_adv;
3600: oepos = nd_epos;
3601: if ( obpe < 2 ) nd_bpe = 2;
3602: else if ( obpe < 3 ) nd_bpe = 3;
3603: else if ( obpe < 4 ) nd_bpe = 4;
3604: else if ( obpe < 5 ) nd_bpe = 5;
3605: else if ( obpe < 6 ) nd_bpe = 6;
3606: else if ( obpe < 8 ) nd_bpe = 8;
3607: else if ( obpe < 10 ) nd_bpe = 10;
3608: else if ( obpe < 16 ) nd_bpe = 16;
3609: else if ( obpe < 32 ) nd_bpe = 32;
3610: else error("nd_reconstruct : exponent too large");
3611:
3612: nd_setup_parameters(nd_nvar,0);
3613: prev_nm_free_list = _nm_free_list;
3614: prev_ndp_free_list = _ndp_free_list;
3615: _nm_free_list = 0;
3616: _ndp_free_list = 0;
3617: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
3618: if ( trace )
3619: for ( i = nd_psn-1; i >= 0; i-- )
3620: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
3621: s0 = 0;
3622: for ( t = d; t; t = NEXT(t) ) {
3623: NEXTND_pairs(s0,s);
3624: s->i1 = t->i1;
3625: s->i2 = t->i2;
3626: SG(s) = SG(t);
3627: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
3628: }
3629:
3630: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
3631: for ( i = 0; i < REDTAB_LEN; i++ ) {
3632: old_red[i] = nd_red[i];
3633: nd_red[i] = 0;
3634: }
3635: for ( i = 0; i < REDTAB_LEN; i++ )
3636: for ( r = old_red[i]; r; r = NEXT(r) ) {
3637: NEWRHist(mr);
3638: mr->index = r->index;
3639: SG(mr) = SG(r);
3640: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
3641: h = ndl_hash_value(DL(mr));
3642: NEXT(mr) = nd_red[h];
3643: nd_red[h] = mr;
3644: }
3645: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
3646: old_red = 0;
3647: for ( i = 0; i < nd_psn; i++ ) {
3648: NEWRHist(r); SG(r) = SG(nd_psh[i]);
3649: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
3650: nd_psh[i] = r;
3651: }
3652: if ( s0 ) NEXT(s) = 0;
3653: prev_nm_free_list = 0;
3654: prev_ndp_free_list = 0;
1.71 noro 3655: #if 0
1.157 noro 3656: GC_gcollect();
1.71 noro 3657: #endif
1.157 noro 3658: return s0;
1.1 noro 3659: }
3660:
1.61 noro 3661: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3662: {
1.157 noro 3663: int n,i,ei,oepw,omask0,j,s,ord_l,l;
3664: struct order_pair *op;
1.1 noro 3665:
1.157 noro 3666: n = nd_nvar;
3667: oepw = (sizeof(UINT)*8)/obpe;
3668: omask0 = (1<<obpe)-1;
3669: TD(r) = TD(d);
3670: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
3671: if ( nd_blockmask ) {
3672: l = nd_blockmask->n;
3673: op = nd_blockmask->order_pair;
3674: for ( i = 1; i < nd_exporigin; i++ )
3675: r[i] = d[i];
3676: for ( j = 0, s = 0; j < l; j++ ) {
3677: ord_l = op[j].length;
3678: for ( i = 0; i < ord_l; i++, s++ ) {
3679: ei = GET_EXP_OLD(d,s);
3680: PUT_EXP(r,s,ei);
3681: }
3682: }
3683: } else {
3684: for ( i = 0; i < n; i++ ) {
3685: ei = GET_EXP_OLD(d,i);
3686: PUT_EXP(r,i,ei);
3687: }
3688: }
3689: if ( nd_module ) MPOS(r) = MPOS(d);
1.1 noro 3690: }
1.3 noro 3691:
1.6 noro 3692: ND nd_copy(ND p)
3693: {
1.157 noro 3694: NM m,mr,mr0;
3695: int c,n;
3696: ND r;
3697:
3698: if ( !p )
3699: return 0;
3700: else {
3701: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3702: NEXTNM(mr0,mr);
3703: CM(mr) = CM(m);
3704: ndl_copy(DL(m),DL(mr));
3705: }
3706: NEXT(mr) = 0;
3707: MKND(NV(p),mr0,LEN(p),r);
3708: SG(r) = SG(p);
3709: return r;
3710: }
1.6 noro 3711: }
3712:
1.53 noro 3713: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3714: {
1.157 noro 3715: NM m1,m2;
3716: NDV p1,p2;
3717: ND t1,t2;
3718: UINT *lcm;
3719: P gp,tp;
3720: Q g,t;
3721: int td;
3722:
3723: if ( !mod && nd_demand ) {
3724: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
3725: } else {
3726: if ( trace ) {
3727: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3728: } else {
3729: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3730: }
3731: }
3732: lcm = LCM(p);
3733: NEWNM(m1); ndl_sub(lcm,HDL(p1),DL(m1));
3734: if ( ndl_check_bound2(p->i1,DL(m1)) ) {
3735: FREENM(m1); return 0;
3736: }
3737: NEWNM(m2); ndl_sub(lcm,HDL(p2),DL(m2));
3738: if ( ndl_check_bound2(p->i2,DL(m2)) ) {
3739: FREENM(m1); FREENM(m2); return 0;
3740: }
3741:
3742: if ( mod == -1 ) {
3743: CM(m1) = HCM(p2); CM(m2) = _chsgnsf(HCM(p1));
3744: } else if ( mod ) {
3745: CM(m1) = HCM(p2); CM(m2) = mod-HCM(p1);
3746: } else if ( nd_vc ) {
3747: ezgcdpz(nd_vc,HCP(p1),HCP(p2),&gp);
3748: divsp(nd_vc,HCP(p2),gp,&CP(m1));
3749: divsp(nd_vc,HCP(p1),gp,&tp); chsgnp(tp,&CP(m2));
3750: } else {
3751: igcd_cofactor(HCQ(p1),HCQ(p2),&g,&t,&CQ(m1)); chsgnq(t,&CQ(m2));
3752: }
3753: t1 = ndv_mul_nm(mod,m1,p1); t2 = ndv_mul_nm(mod,m2,p2);
3754: *rp = nd_add(mod,t1,t2);
3755: FREENM(m1); FREENM(m2);
3756: return 1;
1.11 noro 3757: }
3758:
1.19 noro 3759: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3760: {
1.157 noro 3761: NMV m;
3762: int c,c1,len,i;
1.11 noro 3763:
1.157 noro 3764: if ( !p ) return;
3765: len = LEN(p);
3766: if ( mod == -1 )
3767: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3768: CM(m) = _mulsf(CM(m),mul);
3769: else
3770: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3771: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3772: }
1.11 noro 3773: }
3774:
1.113 noro 3775: void ndv_mul_c_q(NDV p,Q mul)
1.16 noro 3776: {
1.157 noro 3777: NMV m;
3778: Q c;
3779: int len,i;
3780:
3781: if ( !p ) return;
3782: len = LEN(p);
3783: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3784: mulq(CQ(m),mul,&c); CQ(m) = c;
3785: }
1.16 noro 3786: }
3787:
1.55 noro 3788: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
1.157 noro 3789: int n2,i,j,l,n,tlen;
3790: UINT *d0;
3791: NM *tab,*psum;
3792: ND s,r;
3793: NM t;
3794: NMV m1;
3795:
3796: if ( !p ) return 0;
3797: n = NV(p); n2 = n>>1;
3798: d0 = DL(m0);
3799: l = LEN(p);
3800: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3801: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3802: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3803: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
3804: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3805: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
3806: /* m0(NM) * m1(NMV) => tab(NM) */
3807: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
3808: for ( j = 0; j < tlen; j++ ) {
3809: if ( tab[j] ) {
3810: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3811: }
3812: }
3813: }
3814: for ( i = tlen-1, r = 0; i >= 0; i-- )
3815: if ( psum[i] ) {
3816: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3817: MKND(n,psum[i],j,s);
3818: r = nd_add(mod,r,s);
3819: }
3820: if ( r ) SG(r) = SG(p)+TD(d0);
3821: return r;
1.55 noro 3822: }
3823:
1.56 noro 3824: /* product of monomials */
3825: /* XXX block order is not handled correctly */
3826:
1.55 noro 3827: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3828: {
1.157 noro 3829: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
3830: UINT *d0,*d1,*d,*dt,*ctab;
3831: Q *ctab_q;
3832: Q q,q1;
3833: UINT c0,c1,c;
3834: NM *p;
3835: NM m,t;
3836: int mpos;
3837:
3838: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3839: if ( !m0 || !m1 ) return;
3840: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3841: NEWNM(m); d = DL(m);
3842: if ( mod ) {
3843: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3844: } else
3845: mulq(CQ(m0),CQ(m1),&CQ(m));
3846: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3847: homo = n&1 ? 1 : 0;
3848: if ( homo ) {
3849: /* offset of h-degree */
3850: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3851: PUT_EXP(DL(m),n-1,h);
3852: TD(DL(m)) = h;
3853: if ( nd_blockmask ) ndl_weight_mask(DL(m));
3854: }
3855: if ( nd_module ) {
3856: mpos = MPOS(d1);
3857: if ( MPOS(d0) ) error("weyl_mul_nm_nmv : invalid operation");
3858: }
3859: tab[0] = m;
3860: NEWNM(m); d = DL(m);
3861: for ( i = 0, curlen = 1; i < n2; i++ ) {
3862: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3863: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3864: /* xi^a*(Di^k*xi^l)*Di^b */
3865: a += l; b += k;
3866: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
3867: if ( !k || !l ) {
3868: for ( j = 0; j < curlen; j++ )
3869: if ( t = tab[j] ) {
3870: dt = DL(t);
3871: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
3872: if ( nd_blockmask ) ndl_weight_mask(dt);
3873: }
3874: curlen *= k+1;
3875: continue;
3876: }
3877: min = MIN(k,l);
3878: if ( mod ) {
3879: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
3880: mkwcm(k,l,mod,ctab);
3881: } else {
3882: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3883: mkwc(k,l,ctab_q);
3884: }
3885: for ( j = min; j >= 0; j-- ) {
3886: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
3887: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
3888: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
3889: if ( homo ) {
3890: TD(d) = s;
3891: PUT_EXP(d,n-1,s-h);
3892: } else TD(d) = h;
3893: if ( nd_blockmask ) ndl_weight_mask(d);
3894: if ( nd_module ) MPOS(d) = mpos;
3895: if ( mod ) c = ctab[j];
3896: else q = ctab_q[j];
3897: p = tab+curlen*j;
3898: if ( j == 0 ) {
3899: for ( u = 0; u < curlen; u++, p++ ) {
3900: if ( tab[u] ) {
3901: ndl_addto(DL(tab[u]),d);
3902: if ( mod ) {
3903: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3904: } else {
3905: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3906: }
3907: }
3908: }
3909: } else {
3910: for ( u = 0; u < curlen; u++, p++ ) {
3911: if ( tab[u] ) {
3912: NEWNM(t);
3913: ndl_add(DL(tab[u]),d,DL(t));
3914: if ( mod ) {
3915: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3916: } else
3917: mulq(CQ(tab[u]),q,&CQ(t));
3918: *p = t;
3919: }
3920: }
3921: }
3922: }
3923: curlen *= k+1;
3924: }
3925: FREENM(m);
1.55 noro 3926: }
3927:
1.63 noro 3928: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3929: {
1.157 noro 3930: NM mr,mr0;
3931: NMV m;
3932: UINT *d,*dt,*dm;
3933: int c,n,td,i,c1,c2,len;
3934: Q q;
3935: ND r;
3936:
3937: if ( !p ) return 0;
3938: else {
3939: n = NV(p); m = BDY(p);
3940: d = DL(m0);
3941: len = LEN(p);
3942: mr0 = 0;
3943: td = TD(d);
3944: c = CM(m0);
3945: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3946: NEXTNM(mr0,mr);
3947: CM(mr) = 1;
3948: ndl_add(DL(m),d,DL(mr));
3949: }
3950: NEXT(mr) = 0;
3951: MKND(NV(p),mr0,len,r);
3952: SG(r) = SG(p) + TD(d);
3953: return r;
3954: }
1.63 noro 3955: }
3956:
1.55 noro 3957: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3958: {
1.157 noro 3959: NM mr,mr0;
3960: NMV m;
3961: UINT *d,*dt,*dm;
3962: int c,n,td,i,c1,c2,len;
3963: P q;
3964: ND r;
3965:
3966: if ( !p ) return 0;
3967: else if ( do_weyl )
3968: if ( mod == -1 )
3969: error("ndv_mul_nm : not implemented (weyl)");
3970: else
3971: return weyl_ndv_mul_nm(mod,m0,p);
3972: else {
3973: n = NV(p); m = BDY(p);
3974: d = DL(m0);
3975: len = LEN(p);
3976: mr0 = 0;
3977: td = TD(d);
3978: if ( mod == -1 ) {
3979: c = CM(m0);
3980: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3981: NEXTNM(mr0,mr);
3982: CM(mr) = _mulsf(CM(m),c);
3983: ndl_add(DL(m),d,DL(mr));
3984: }
3985: } else if ( mod ) {
3986: c = CM(m0);
3987: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3988: NEXTNM(mr0,mr);
3989: c1 = CM(m);
3990: DMAR(c1,c,0,mod,c2);
3991: CM(mr) = c2;
3992: ndl_add(DL(m),d,DL(mr));
3993: }
3994: } else {
3995: q = CP(m0);
3996: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3997: NEXTNM(mr0,mr);
3998: mulp(nd_vc,CP(m),q,&CP(mr));
3999: ndl_add(DL(m),d,DL(mr));
4000: }
4001: }
4002: NEXT(mr) = 0;
4003: MKND(NV(p),mr0,len,r);
4004: SG(r) = SG(p) + TD(d);
4005: return r;
4006: }
1.4 noro 4007: }
4008:
1.104 noro 4009: ND nd_quo(int mod,PGeoBucket bucket,NDV d)
1.99 noro 4010: {
1.157 noro 4011: NM mq0,mq;
4012: NMV tm;
4013: Q q;
4014: int i,nv,sg,c,c1,c2,hindex;
4015: ND p,t,r;
4016: N tnm;
4017:
4018: if ( bucket->m < 0 ) return 0;
4019: else {
4020: nv = NV(d);
4021: mq0 = 0;
4022: tm = (NMV)ALLOCA(nmv_adv);
4023: while ( 1 ) {
4024: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
4025: if ( hindex < 0 ) break;
4026: p = bucket->body[hindex];
4027: NEXTNM(mq0,mq);
4028: ndl_sub(HDL(p),HDL(d),DL(mq));
4029: ndl_copy(DL(mq),DL(tm));
4030: if ( mod ) {
4031: c1 = invm(HCM(d),mod); c2 = HCM(p);
4032: DMAR(c1,c2,0,mod,c); CM(mq) = c;
4033: CM(tm) = mod-c;
4034: } else {
4035: divsn(NM(HCQ(p)),NM(HCQ(d)),&tnm);
4036: NTOQ(tnm,SGN(HCQ(p))*SGN(HCQ(d)),CQ(mq));
4037: chsgnq(CQ(mq),&CQ(tm));
4038: }
4039: t = ndv_mul_nmv_trunc(mod,tm,d,HDL(d));
4040: bucket->body[hindex] = nd_remove_head(p);
4041: t = nd_remove_head(t);
4042: add_pbucket(mod,bucket,t);
4043: }
4044: if ( !mq0 )
4045: r = 0;
4046: else {
4047: NEXT(mq) = 0;
4048: for ( i = 0, mq = mq0; mq; mq = NEXT(mq), i++ );
4049: MKND(nv,mq0,i,r);
4050: /* XXX */
4051: SG(r) = HTD(r);
4052: }
4053: return r;
4054: }
1.99 noro 4055: }
4056:
1.43 noro 4057: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 4058: {
1.157 noro 4059: NMV m,mr,mr0,t;
4060: int len,i,k;
1.11 noro 4061:
1.157 noro 4062: if ( !p ) return;
4063: m = BDY(p); len = LEN(p);
4064: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
4065: m = (NMV)((char *)mr0+(len-1)*oadv);
4066: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
4067: t = (NMV)ALLOCA(nmv_adv);
4068: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
4069: CQ(t) = CQ(m);
4070: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
4071: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
4072: CQ(mr) = CQ(t);
4073: ndl_copy(DL(t),DL(mr));
4074: }
4075: BDY(p) = mr0;
1.61 noro 4076: }
4077:
4078: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
4079: {
1.157 noro 4080: NMV m,mr,mr0;
4081: int len,i;
4082: NDV r;
4083:
4084: if ( !p ) return 0;
4085: m = BDY(p); len = LEN(p);
4086: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
4087: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
4088: ndl_zero(DL(mr));
4089: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
4090: CQ(mr) = CQ(m);
4091: }
4092: MKNDV(NV(p),mr0,len,r);
4093: SG(r) = SG(p);
4094: return r;
1.11 noro 4095: }
4096:
1.61 noro 4097: /* duplicate p */
4098:
4099: NDV ndv_dup(int mod,NDV p)
1.3 noro 4100: {
1.157 noro 4101: NDV d;
4102: NMV t,m,m0;
4103: int i,len;
4104:
4105: if ( !p ) return 0;
4106: len = LEN(p);
4107: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
4108: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
4109: ndl_copy(DL(t),DL(m));
4110: CQ(m) = CQ(t);
4111: }
4112: MKNDV(NV(p),m0,len,d);
4113: SG(d) = SG(p);
4114: return d;
1.23 noro 4115: }
4116:
1.63 noro 4117: ND nd_dup(ND p)
4118: {
1.157 noro 4119: ND d;
4120: NM t,m,m0;
1.63 noro 4121:
1.157 noro 4122: if ( !p ) return 0;
4123: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
4124: NEXTNM(m0,m);
4125: ndl_copy(DL(t),DL(m));
4126: CQ(m) = CQ(t);
4127: }
4128: if ( m0 ) NEXT(m) = 0;
4129: MKND(NV(p),m0,LEN(p),d);
4130: SG(d) = SG(p);
4131: return d;
1.63 noro 4132: }
4133:
1.61 noro 4134: /* XXX if p->len == 0 then it represents 0 */
4135:
4136: void ndv_mod(int mod,NDV p)
4137: {
1.157 noro 4138: NMV t,d;
4139: int r,s,u;
4140: int i,len,dlen;
4141: P cp;
4142: Q c;
4143: Obj gfs;
4144:
4145: if ( !p ) return;
4146: len = LEN(p);
4147: dlen = 0;
4148: if ( mod == -1 )
4149: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
4150: simp_ff((Obj)CP(t),&gfs);
4151: r = FTOIF(CONT((GFS)gfs));
4152: CM(d) = r;
4153: ndl_copy(DL(t),DL(d));
4154: NMV_ADV(d);
4155: dlen++;
4156: }
4157: else
4158: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
4159: if ( nd_vc ) {
4160: nd_subst_vector(nd_vc,CP(t),nd_subst,&cp);
4161: c = (Q)cp;
4162: } else
4163: c = CQ(t);
4164: r = rem(NM(c),mod);
4165: if ( r ) {
4166: if ( SGN(c) < 0 )
4167: r = mod-r;
4168: if ( DN(c) ) {
4169: s = rem(DN(c),mod);
4170: if ( !s )
4171: error("ndv_mod : division by 0");
4172: s = invm(s,mod);
4173: DMAR(r,s,0,mod,u); r = u;
4174: }
4175: CM(d) = r;
4176: ndl_copy(DL(t),DL(d));
4177: NMV_ADV(d);
4178: dlen++;
4179: }
4180: }
4181: LEN(p) = dlen;
1.61 noro 4182: }
4183:
4184: NDV ptondv(VL vl,VL dvl,P p)
4185: {
1.157 noro 4186: ND nd;
4187:
4188: nd = ptond(vl,dvl,p);
4189: return ndtondv(0,nd);
4190: }
1.61 noro 4191:
1.157 noro 4192: void pltozpl(LIST l,Q *cont,LIST *pp)
4193: {
4194: NODE nd,nd1;
4195: int n;
4196: P *pl;
4197: Q *cl;
4198: int i;
4199: P dmy;
4200: Q dvr;
4201: LIST r;
4202:
4203: nd = BDY(l); n = length(nd);
4204: pl = (P *)ALLOCA(n*sizeof(P));
4205: cl = (Q *)ALLOCA(n*sizeof(P));
4206: for ( i = 0; i < n; i++, nd = NEXT(nd) )
4207: ptozp((P)BDY(nd),1,&cl[i],&dmy);
4208: qltozl(cl,n,&dvr);
4209: nd = BDY(l);
4210: for ( i = 0; i < n; i++, nd = NEXT(nd) ) {
4211: divsp(CO,(P)BDY(nd),(P)dvr,&pl[i]);
4212: }
4213: nd = 0;
4214: for ( i = n-1; i >= 0; i-- ) {
4215: MKNODE(nd1,pl[i],nd); nd = nd1;
4216: }
4217: MKLIST(r,nd);
4218: *pp = r;
4219: }
4220:
4221: /* (a1,a2,...,an) -> a1*e(1)+...+an*e(n) */
4222:
4223: NDV pltondv(VL vl,VL dvl,LIST p)
4224: {
4225: int i;
4226: NODE t;
4227: ND r,ri;
4228: NM m;
4229:
4230: if ( !nd_module ) error("pltond : module order must be set");
4231: r = 0;
4232: for ( i = 1, t = BDY(p); t; t = NEXT(t), i++ ) {
4233: ri = ptond(vl,dvl,(P)BDY(t));
4234: if ( ri )
4235: for ( m = BDY(ri); m; m = NEXT(m) )
4236: MPOS(DL(m)) = i;
4237: r = nd_add(0,r,ri);
4238: }
4239: return ndtondv(0,r);
1.61 noro 4240: }
4241:
4242: ND ptond(VL vl,VL dvl,P p)
1.23 noro 4243: {
1.157 noro 4244: int n,i,j,k,e;
4245: VL tvl;
4246: V v;
4247: DCP dc;
4248: DCP *w;
4249: ND r,s,t,u;
4250: P x;
4251: int c;
4252: UINT *d;
4253: NM m,m0;
4254:
4255: if ( !p )
4256: return 0;
4257: else if ( NUM(p) ) {
4258: NEWNM(m);
4259: ndl_zero(DL(m));
4260: CQ(m) = (Q)p;
4261: NEXT(m) = 0;
4262: MKND(nd_nvar,m,1,r);
4263: SG(r) = 0;
4264: return r;
4265: } else {
4266: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
4267: w = (DCP *)ALLOCA(k*sizeof(DCP));
4268: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
4269: for ( i = 0, tvl = dvl, v = VR(p);
4270: tvl && tvl->v != v; tvl = NEXT(tvl), i++ );
4271: if ( !tvl ) {
4272: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
4273: t = ptond(vl,dvl,COEF(w[j]));
4274: pwrp(vl,x,DEG(w[j]),&p);
4275: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
4276: }
4277: return s;
4278: } else {
4279: NEWNM(m0); d = DL(m0);
4280: for ( j = k-1, s = 0; j >= 0; j-- ) {
4281: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
4282: TD(d) = MUL_WEIGHT(e,i);
4283: if ( nd_blockmask) ndl_weight_mask(d);
4284: if ( nd_module ) MPOS(d) = 0;
4285: t = ptond(vl,dvl,COEF(w[j]));
4286: for ( m = BDY(t); m; m = NEXT(m) )
4287: ndl_addto(DL(m),d);
4288: SG(t) += TD(d);
4289: s = nd_add(0,s,t);
4290: }
4291: FREENM(m0);
4292: return s;
4293: }
4294: }
1.61 noro 4295: }
4296:
4297: P ndvtop(int mod,VL vl,VL dvl,NDV p)
4298: {
1.157 noro 4299: VL tvl;
4300: int len,n,j,i,e;
4301: NMV m;
4302: Q q;
4303: P c;
4304: UINT *d;
4305: P s,r,u,t,w;
4306: GFS gfs;
4307:
4308: if ( !p ) return 0;
4309: else {
4310: len = LEN(p);
4311: n = NV(p);
4312: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
4313: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
4314: if ( mod == -1 ) {
4315: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
4316: } else if ( mod ) {
4317: STOQ(CM(m),q); c = (P)q;
4318: } else
4319: c = CP(m);
4320: d = DL(m);
4321: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
4322: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
4323: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
4324: }
4325: addp(vl,s,t,&u); s = u;
4326: }
4327: return s;
4328: }
4329: }
4330:
4331: LIST ndvtopl(int mod,VL vl,VL dvl,NDV p,int rank)
4332: {
4333: VL tvl;
4334: int len,n,j,i,e;
4335: NMV m;
4336: Q q;
4337: P c;
4338: UINT *d;
4339: P s,r,u,t,w;
4340: GFS gfs;
4341: P *a;
4342: LIST l;
4343: NODE nd,nd1;
4344:
4345: if ( !p ) return 0;
4346: else {
4347: a = (P *)ALLOCA((rank+1)*sizeof(P));
4348: for ( i = 0; i <= rank; i++ ) a[i] = 0;
4349: len = LEN(p);
4350: n = NV(p);
4351: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
4352: for ( j = len-1; j >= 0; j--, NMV_PREV(m) ) {
4353: if ( mod == -1 ) {
4354: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
4355: } else if ( mod ) {
4356: STOQ(CM(m),q); c = (P)q;
4357: } else
4358: c = CP(m);
4359: d = DL(m);
4360: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
4361: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
4362: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
4363: }
4364: addp(vl,a[MPOS(d)],t,&u); a[MPOS(d)] = u;
4365: }
4366: nd = 0;
4367: for ( i = rank; i > 0; i-- ) {
4368: MKNODE(nd1,a[i],nd); nd = nd1;
4369: }
4370: MKLIST(l,nd);
4371: return l;
4372: }
1.3 noro 4373: }
4374:
1.61 noro 4375: NDV ndtondv(int mod,ND p)
1.11 noro 4376: {
1.157 noro 4377: NDV d;
4378: NMV m,m0;
4379: NM t;
4380: int i,len;
4381:
4382: if ( !p ) return 0;
4383: len = LEN(p);
4384: if ( mod )
4385: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
4386: else
4387: m0 = m = MALLOC(len*nmv_adv);
1.103 noro 4388: #if 0
1.157 noro 4389: ndv_alloc += nmv_adv*len;
1.103 noro 4390: #endif
1.157 noro 4391: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
4392: ndl_copy(DL(t),DL(m));
4393: CQ(m) = CQ(t);
4394: }
4395: MKNDV(NV(p),m0,len,d);
4396: SG(d) = SG(p);
4397: return d;
1.11 noro 4398: }
4399:
1.61 noro 4400: ND ndvtond(int mod,NDV p)
1.11 noro 4401: {
1.157 noro 4402: ND d;
4403: NM m,m0;
4404: NMV t;
4405: int i,len;
4406:
4407: if ( !p ) return 0;
4408: m0 = 0;
4409: len = p->len;
4410: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
4411: NEXTNM(m0,m);
4412: ndl_copy(DL(t),DL(m));
4413: CQ(m) = CQ(t);
4414: }
4415: NEXT(m) = 0;
4416: MKND(NV(p),m0,len,d);
4417: SG(d) = SG(p);
4418: return d;
1.11 noro 4419: }
4420:
1.3 noro 4421: void ndv_print(NDV p)
4422: {
1.157 noro 4423: NMV m;
4424: int i,len;
1.3 noro 4425:
1.157 noro 4426: if ( !p ) printf("0\n");
4427: else {
4428: len = LEN(p);
4429: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4430: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
4431: else printf("+%d*",CM(m));
4432: ndl_print(DL(m));
4433: }
4434: printf("\n");
4435: }
1.16 noro 4436: }
4437:
1.113 noro 4438: void ndv_print_q(NDV p)
1.16 noro 4439: {
1.157 noro 4440: NMV m;
4441: int i,len;
1.16 noro 4442:
1.157 noro 4443: if ( !p ) printf("0\n");
4444: else {
4445: len = LEN(p);
4446: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4447: printf("+");
4448: printexpr(CO,(Obj)CQ(m));
4449: printf("*");
4450: ndl_print(DL(m));
4451: }
4452: printf("\n");
4453: }
1.25 noro 4454: }
4455:
1.61 noro 4456: NODE ndv_reducebase(NODE x)
1.27 noro 4457: {
1.157 noro 4458: int len,i,j;
4459: NDV *w;
4460: NODE t,t0;
4461:
4462: len = length(x);
4463: w = (NDV *)ALLOCA(len*sizeof(NDV));
4464: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
4465: for ( i = 0; i < len; i++ ) {
4466: for ( j = 0; j < i; j++ ) {
4467: if ( w[i] && w[j] )
4468: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
4469: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
4470: }
4471: }
4472: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
4473: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
4474: }
4475: NEXT(t) = 0; x = t0;
4476: return x;
1.11 noro 4477: }
1.32 noro 4478:
1.43 noro 4479: /* XXX incomplete */
4480:
1.32 noro 4481: void nd_init_ord(struct order_spec *ord)
4482: {
1.157 noro 4483: nd_module = (ord->id >= 256);
4484: switch ( ord->id ) {
4485: case 0:
4486: switch ( ord->ord.simple ) {
4487: case 0:
4488: nd_dcomp = 1;
4489: nd_isrlex = 1;
4490: break;
4491: case 1:
4492: nd_dcomp = 1;
4493: nd_isrlex = 0;
4494: break;
4495: case 2:
4496: nd_dcomp = 0;
4497: nd_isrlex = 0;
4498: ndl_compare_function = ndl_lex_compare;
4499: break;
4500: case 11:
4501: /* XXX */
4502: nd_dcomp = 0;
4503: nd_isrlex = 1;
4504: ndl_compare_function = ndl_ww_lex_compare;
4505: break;
4506: default:
4507: error("nd_gr : unsupported order");
4508: }
4509: break;
4510: case 1:
4511: /* block order */
4512: /* XXX */
4513: nd_dcomp = -1;
4514: nd_isrlex = 0;
4515: ndl_compare_function = ndl_block_compare;
4516: break;
4517: case 2:
4518: /* matrix order */
4519: /* XXX */
4520: nd_dcomp = -1;
4521: nd_isrlex = 0;
4522: nd_matrix_len = ord->ord.matrix.row;
4523: nd_matrix = ord->ord.matrix.matrix;
4524: ndl_compare_function = ndl_matrix_compare;
4525: break;
4526: case 3:
4527: /* composite order */
4528: nd_dcomp = -1;
4529: nd_isrlex = 0;
4530: nd_worb_len = ord->ord.composite.length;
4531: nd_worb = ord->ord.composite.w_or_b;
4532: ndl_compare_function = ndl_composite_compare;
4533: break;
4534:
4535: /* module order */
4536: case 256:
4537: nd_istop = ord->istop;
4538: nd_dcomp = -1;
4539: nd_isrlex = 0;
4540: switch ( ord->ord.simple ) {
4541: case 0:
4542: ndl_compare_function = ndl_module_grlex_compare;
4543: break;
4544: case 1:
4545: ndl_compare_function = ndl_module_glex_compare;
4546: break;
4547: case 2:
4548: ndl_compare_function = ndl_module_lex_compare;
4549: break;
4550: default:
4551: error("nd_gr : unsupported order");
4552: }
4553: break;
4554: case 257:
4555: /* block order */
4556: ndl_compare_function = ndl_module_block_compare;
4557: break;
4558: case 258:
4559: /* matrix order */
4560: nd_matrix_len = ord->ord.matrix.row;
4561: nd_matrix = ord->ord.matrix.matrix;
4562: ndl_compare_function = ndl_module_matrix_compare;
4563: break;
4564: case 259:
4565: /* composite order */
4566: nd_worb_len = ord->ord.composite.length;
4567: nd_worb = ord->ord.composite.w_or_b;
4568: ndl_compare_function = ndl_module_composite_compare;
4569: break;
4570: }
4571: nd_ord = ord;
1.32 noro 4572: }
4573:
1.43 noro 4574: BlockMask nd_create_blockmask(struct order_spec *ord)
4575: {
1.157 noro 4576: int n,i,j,s,l;
4577: UINT *t;
4578: BlockMask bm;
4579:
4580: /* we only create mask table for block order */
4581: if ( ord->id != 1 )
4582: return 0;
4583: n = ord->ord.block.length;
4584: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
4585: bm->n = n;
4586: bm->order_pair = ord->ord.block.order_pair;
4587: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
4588: for ( i = 0, s = 0; i < n; i++ ) {
4589: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
4590: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
4591: l = bm->order_pair[i].length;
4592: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
4593: }
4594: return bm;
1.57 noro 4595: }
4596:
4597: EPOS nd_create_epos(struct order_spec *ord)
4598: {
1.157 noro 4599: int i,j,l,s,ord_l,ord_o;
4600: EPOS epos;
4601: struct order_pair *op;
4602:
4603: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
4604: switch ( ord->id ) {
4605: case 0:
4606: if ( nd_isrlex ) {
4607: for ( i = 0; i < nd_nvar; i++ ) {
4608: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
4609: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
4610: }
4611: } else {
4612: for ( i = 0; i < nd_nvar; i++ ) {
4613: epos[i].i = nd_exporigin + i/nd_epw;
4614: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4615: }
4616: }
4617: break;
4618: case 1:
4619: /* block order */
4620: l = ord->ord.block.length;
4621: op = ord->ord.block.order_pair;
4622: for ( j = 0, s = 0; j < l; j++ ) {
4623: ord_o = op[j].order;
4624: ord_l = op[j].length;
4625: if ( !ord_o )
4626: for ( i = 0; i < ord_l; i++ ) {
4627: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
4628: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
4629: }
4630: else
4631: for ( i = 0; i < ord_l; i++ ) {
4632: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
4633: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
4634: }
4635: s += ord_l;
4636: }
4637: break;
4638: case 2:
4639: /* matrix order */
4640: case 3:
4641: /* composite order */
4642: default:
4643: for ( i = 0; i < nd_nvar; i++ ) {
4644: epos[i].i = nd_exporigin + i/nd_epw;
4645: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
4646: }
4647: break;
4648: }
4649: return epos;
1.43 noro 4650: }
1.59 noro 4651:
4652: /* external interface */
4653:
4654: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
4655: {
1.157 noro 4656: NODE t,in0,in;
4657: ND nd,nf;
4658: NDV ndv;
4659: VL vv,tv;
4660: int stat,nvar,max,e;
4661: union oNDC dn;
4662: Q cont;
4663: P pp;
4664:
4665: if ( !f ) {
4666: *rp = 0;
4667: return;
4668: }
4669: pltovl(v,&vv);
4670: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
4671:
4672: /* get the degree bound */
4673: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
4674: for ( tv = vv; tv; tv = NEXT(tv) ) {
4675: e = getdeg(tv->v,(P)BDY(t));
4676: max = MAX(e,max);
4677: }
4678: for ( tv = vv; tv; tv = NEXT(tv) ) {
4679: e = getdeg(tv->v,f);
4680: max = MAX(e,max);
4681: }
4682:
4683: nd_init_ord(ord);
4684: nd_setup_parameters(nvar,max);
4685:
4686: /* conversion to ndv */
4687: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
4688: NEXTNODE(in0,in);
4689: ptozp((P)BDY(t),1,&cont,&pp);
4690: BDY(in) = (pointer)ptondv(CO,vv,pp);
4691: if ( m ) ndv_mod(m,(NDV)BDY(in));
4692: }
4693: NEXTNODE(in0,in);
4694: BDY(in) = (pointer)ptondv(CO,vv,f);
4695: if ( m ) ndv_mod(m,(NDV)BDY(in));
4696: NEXT(in) = 0;
4697:
4698: /* dont sort, dont removecont */
4699: ndv_setup(m,0,in0,1,1);
4700: nd_psn--;
4701: nd_scale=2;
4702: while ( 1 ) {
4703: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
4704: stat = nd_nf(m,0,nd,nd_ps,1,0,&nf);
4705: if ( !stat ) {
4706: nd_psn++;
4707: nd_reconstruct(0,0);
4708: nd_psn--;
4709: } else
4710: break;
4711: }
4712: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 4713: }
4714:
4715: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
4716: {
1.157 noro 4717: NM m;
4718: UINT *t,*s;
4719: int i;
4720:
4721: for ( i = 0; i < n; i++ ) r[i] = 0;
4722: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4723: t = DL(m);
4724: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4725: r[i] = CM(m);
4726: }
4727: for ( i = 0; !r[i]; i++ );
4728: return i;
1.63 noro 4729: }
4730:
1.113 noro 4731: int nd_to_vect_q(UINT *s0,int n,ND d,Q *r)
1.74 noro 4732: {
1.157 noro 4733: NM m;
4734: UINT *t,*s;
4735: int i;
4736:
4737: for ( i = 0; i < n; i++ ) r[i] = 0;
4738: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
4739: t = DL(m);
4740: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4741: r[i] = CQ(m);
4742: }
4743: for ( i = 0; !r[i]; i++ );
4744: return i;
1.74 noro 4745: }
4746:
1.129 noro 4747: Q *nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair)
4748: {
1.157 noro 4749: NM m;
4750: NMV mr;
4751: UINT *d,*t,*s;
4752: NDV p;
4753: int i,j,len;
4754: Q *r;
4755:
4756: m = pair->mul;
4757: d = DL(m);
4758: p = nd_ps[pair->index];
4759: len = LEN(p);
4760: r = (Q *)CALLOC(n,sizeof(Q));
4761: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4762: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4763: ndl_add(d,DL(mr),t);
4764: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4765: r[i] = CQ(mr);
4766: }
4767: return r;
1.129 noro 4768: }
4769:
1.67 noro 4770: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 4771: {
1.157 noro 4772: NM m;
4773: NMV mr;
4774: UINT *d,*t,*s;
4775: NDV p;
4776: unsigned char *ivc;
4777: unsigned short *ivs;
4778: UINT *v,*ivi,*s0v;
4779: int i,j,len,prev,diff,cdiff;
4780: IndArray r;
4781:
4782: m = pair->mul;
4783: d = DL(m);
4784: p = nd_ps[pair->index];
4785: len = LEN(p);
4786: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
4787: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
4788: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
4789: ndl_add(d,DL(mr),t);
4790: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
4791: v[j] = i;
4792: }
4793: r = (IndArray)MALLOC(sizeof(struct oIndArray));
4794: r->head = v[0];
4795: diff = 0;
4796: for ( i = 1; i < len; i++ ) {
4797: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
4798: }
4799: if ( diff < 256 ) {
4800: r->width = 1;
4801: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
4802: r->index.c = ivc;
4803: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
4804: } else if ( diff < 65536 ) {
4805: r->width = 2;
4806: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
4807: r->index.s = ivs;
4808: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
4809: } else {
4810: r->width = 4;
4811: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
4812: r->index.i = ivi;
4813: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
4814: }
4815: return r;
1.64 noro 4816: }
4817:
1.135 noro 4818: int compress_array(Q *svect,Q *cvect,int n)
4819: {
1.157 noro 4820: int i,j;
1.135 noro 4821:
1.157 noro 4822: for ( i = j = 0; i < n; i++ )
4823: if ( svect[i] ) cvect[j++] = svect[i];
4824: return j;
1.135 noro 4825: }
4826:
4827: void expand_array(Q *svect,Q *cvect,int n)
4828: {
1.157 noro 4829: int i,j;
1.135 noro 4830:
1.157 noro 4831: for ( i = j = 0; j < n; i++ )
4832: if ( svect[i] ) svect[i] = cvect[j++];
1.135 noro 4833: }
4834:
1.133 noro 4835: int ndv_reduce_vect_q(Q *svect,int trace,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.107 noro 4836: {
1.157 noro 4837: int i,j,k,len,pos,prev,nz;
4838: Q cs,mcs,c1,c2,cr,gcd,t;
4839: IndArray ivect;
4840: unsigned char *ivc;
4841: unsigned short *ivs;
4842: unsigned int *ivi;
4843: NDV redv;
4844: NMV mr;
4845: NODE rp;
4846: int maxrs;
4847: double hmag;
4848: Q *cvect;
4849:
4850: maxrs = 0;
4851: for ( i = 0; i < col && !svect[i]; i++ );
4852: if ( i == col ) return maxrs;
4853: hmag = p_mag((P)svect[i])*nd_scale;
4854: cvect = (Q *)ALLOCA(col*sizeof(Q));
4855: for ( i = 0; i < nred; i++ ) {
4856: ivect = imat[i];
4857: k = ivect->head;
4858: if ( svect[k] ) {
4859: maxrs = MAX(maxrs,rp0[i]->sugar);
4860: redv = trace?nd_ps_trace[rp0[i]->index]:nd_ps[rp0[i]->index];
4861: len = LEN(redv); mr = BDY(redv);
4862: igcd_cofactor(svect[k],CQ(mr),&gcd,&cs,&cr);
4863: chsgnq(cs,&mcs);
4864: if ( !UNIQ(cr) ) {
4865: for ( j = 0; j < col; j++ ) {
4866: mulq(svect[j],cr,&c1); svect[j] = c1;
4867: }
4868: }
4869: svect[k] = 0; prev = k;
4870: switch ( ivect->width ) {
4871: case 1:
4872: ivc = ivect->index.c;
4873: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4874: pos = prev+ivc[j]; prev = pos;
4875: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
4876: }
4877: break;
4878: case 2:
4879: ivs = ivect->index.s;
4880: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4881: pos = prev+ivs[j]; prev = pos;
4882: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
4883: }
4884: break;
4885: case 4:
4886: ivi = ivect->index.i;
4887: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4888: pos = prev+ivi[j]; prev = pos;
4889: mulq(CQ(mr),mcs,&c2); addq(svect[pos],c2,&t); svect[pos] = t;
4890: }
4891: break;
4892: }
4893: for ( j = k+1; j < col && !svect[j]; j++ );
4894: if ( j == col ) break;
4895: if ( hmag && ((double)p_mag((P)svect[j]) > hmag) ) {
4896: nz = compress_array(svect,cvect,col);
4897: removecont_array((P *)cvect,nz,1);
4898: expand_array(svect,cvect,nz);
4899: hmag = ((double)p_mag((P)svect[j]))*nd_scale;
4900: }
4901: }
4902: }
4903: nz = compress_array(svect,cvect,col);
4904: removecont_array((P *)cvect,nz,1);
4905: expand_array(svect,cvect,nz);
4906: if ( DP_Print ) {
4907: fprintf(asir_out,"-"); fflush(asir_out);
4908: }
4909: return maxrs;
1.107 noro 4910: }
4911:
1.76 noro 4912: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 4913: {
1.157 noro 4914: int i,j,k,len,pos,prev;
4915: UINT c,c1,c2,c3,up,lo,dmy;
4916: IndArray ivect;
4917: unsigned char *ivc;
4918: unsigned short *ivs;
4919: unsigned int *ivi;
4920: NDV redv;
4921: NMV mr;
4922: NODE rp;
4923: int maxrs;
4924:
4925: maxrs = 0;
4926: for ( i = 0; i < nred; i++ ) {
4927: ivect = imat[i];
4928: k = ivect->head; svect[k] %= m;
4929: if ( c = svect[k] ) {
4930: maxrs = MAX(maxrs,rp0[i]->sugar);
4931: c = m-c; redv = nd_ps[rp0[i]->index];
4932: len = LEN(redv); mr = BDY(redv);
4933: svect[k] = 0; prev = k;
4934: switch ( ivect->width ) {
4935: case 1:
4936: ivc = ivect->index.c;
4937: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4938: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
4939: prev = pos;
4940: DMA(c1,c,c2,up,lo);
4941: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4942: } else svect[pos] = lo;
4943: }
4944: break;
4945: case 2:
4946: ivs = ivect->index.s;
4947: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4948: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
4949: prev = pos;
4950: DMA(c1,c,c2,up,lo);
4951: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4952: } else svect[pos] = lo;
4953: }
4954: break;
4955: case 4:
4956: ivi = ivect->index.i;
4957: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4958: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
4959: prev = pos;
4960: DMA(c1,c,c2,up,lo);
4961: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
4962: } else svect[pos] = lo;
4963: }
4964: break;
4965: }
4966: }
4967: }
4968: for ( i = 0; i < col; i++ )
4969: if ( svect[i] >= (UINT)m ) svect[i] %= m;
4970: return maxrs;
1.65 noro 4971: }
4972:
1.76 noro 4973: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 4974: {
1.157 noro 4975: int i,j,k,len,pos,prev;
4976: UINT c,c1,c2,c3,up,lo,dmy;
4977: IndArray ivect;
4978: unsigned char *ivc;
4979: unsigned short *ivs;
4980: unsigned int *ivi;
4981: NDV redv;
4982: NMV mr;
4983: NODE rp;
4984: int maxrs;
4985:
4986: maxrs = 0;
4987: for ( i = 0; i < nred; i++ ) {
4988: ivect = imat[i];
4989: k = ivect->head; svect[k] %= m;
4990: if ( c = svect[k] ) {
4991: maxrs = MAX(maxrs,rp0[i]->sugar);
4992: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
4993: len = LEN(redv); mr = BDY(redv);
4994: svect[k] = 0; prev = k;
4995: switch ( ivect->width ) {
4996: case 1:
4997: ivc = ivect->index.c;
4998: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4999: pos = prev+ivc[j]; prev = pos;
5000: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
5001: }
5002: break;
5003: case 2:
5004: ivs = ivect->index.s;
5005: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
5006: pos = prev+ivs[j]; prev = pos;
5007: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
5008: }
5009: break;
5010: case 4:
5011: ivi = ivect->index.i;
5012: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
5013: pos = prev+ivi[j]; prev = pos;
5014: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
5015: }
5016: break;
5017: }
5018: }
5019: }
5020: return maxrs;
1.72 noro 5021: }
5022:
1.65 noro 5023: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
5024: {
1.157 noro 5025: int j,k,len;
5026: UINT *p;
5027: UINT c;
5028: NDV r;
5029: NMV mr0,mr;
5030:
5031: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
5032: if ( !len ) return 0;
5033: else {
5034: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5035: #if 0
1.157 noro 5036: ndv_alloc += nmv_adv*len;
1.103 noro 5037: #endif
1.157 noro 5038: mr = mr0;
5039: p = s0vect;
5040: for ( j = k = 0; j < col; j++, p += nd_wpd )
5041: if ( !rhead[j] ) {
5042: if ( c = vect[k++] ) {
5043: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
5044: }
5045: }
5046: MKNDV(nd_nvar,mr0,len,r);
5047: return r;
5048: }
1.65 noro 5049: }
5050:
1.129 noro 5051: /* for preprocessed vector */
5052:
1.113 noro 5053: NDV vect_to_ndv_q(Q *vect,int spcol,int col,int *rhead,UINT *s0vect)
1.107 noro 5054: {
1.157 noro 5055: int j,k,len;
5056: UINT *p;
5057: Q c;
5058: NDV r;
5059: NMV mr0,mr;
5060:
5061: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
5062: if ( !len ) return 0;
5063: else {
5064: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.107 noro 5065: #if 0
1.157 noro 5066: ndv_alloc += nmv_adv*len;
1.107 noro 5067: #endif
1.157 noro 5068: mr = mr0;
5069: p = s0vect;
5070: for ( j = k = 0; j < col; j++, p += nd_wpd )
5071: if ( !rhead[j] ) {
5072: if ( c = vect[k++] ) {
5073: if ( DN(c) )
5074: error("afo");
5075: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
5076: }
5077: }
5078: MKNDV(nd_nvar,mr0,len,r);
5079: return r;
5080: }
1.107 noro 5081: }
5082:
1.129 noro 5083: /* for plain vector */
5084:
5085: NDV plain_vect_to_ndv_q(Q *vect,int col,UINT *s0vect)
5086: {
1.157 noro 5087: int j,k,len;
5088: UINT *p;
5089: Q c;
5090: NDV r;
5091: NMV mr0,mr;
5092:
5093: for ( j = 0, len = 0; j < col; j++ ) if ( vect[j] ) len++;
5094: if ( !len ) return 0;
5095: else {
5096: mr0 = (NMV)GC_malloc(nmv_adv*len);
1.129 noro 5097: #if 0
1.157 noro 5098: ndv_alloc += nmv_adv*len;
1.129 noro 5099: #endif
1.157 noro 5100: mr = mr0;
5101: p = s0vect;
5102: for ( j = k = 0; j < col; j++, p += nd_wpd, k++ )
5103: if ( c = vect[k] ) {
5104: if ( DN(c) )
5105: error("afo");
5106: ndl_copy(p,DL(mr)); CQ(mr) = c; NMV_ADV(mr);
5107: }
5108: MKNDV(nd_nvar,mr0,len,r);
5109: return r;
5110: }
1.129 noro 5111: }
5112:
1.133 noro 5113: int nd_sp_f4(int m,int trace,ND_pairs l,PGeoBucket bucket)
1.65 noro 5114: {
1.157 noro 5115: ND_pairs t;
5116: NODE sp0,sp;
5117: int stat;
5118: ND spol;
5119:
5120: for ( t = l; t; t = NEXT(t) ) {
5121: stat = nd_sp(m,trace,t,&spol);
5122: if ( !stat ) return 0;
5123: if ( spol ) {
5124: add_pbucket_symbolic(bucket,spol);
5125: }
5126: }
5127: return 1;
1.65 noro 5128: }
5129:
1.133 noro 5130: int nd_symbolic_preproc(PGeoBucket bucket,int trace,UINT **s0vect,NODE *r)
1.65 noro 5131: {
1.157 noro 5132: NODE rp0,rp;
5133: NM mul,head,s0,s;
5134: int index,col,i,sugar;
5135: RHist h;
5136: UINT *s0v,*p;
5137: NM_ind_pair pair;
5138: ND red;
5139: NDV *ps;
5140:
5141: s0 = 0; rp0 = 0; col = 0;
5142: ps = trace?nd_ps_trace:nd_ps;
5143: while ( 1 ) {
5144: head = remove_head_pbucket_symbolic(bucket);
5145: if ( !head ) break;
5146: if ( !s0 ) s0 = head;
5147: else NEXT(s) = head;
5148: s = head;
5149: index = ndl_find_reducer(DL(head));
5150: if ( index >= 0 ) {
5151: h = nd_psh[index];
5152: NEWNM(mul);
5153: ndl_sub(DL(head),DL(h),DL(mul));
5154: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
5155: sugar = TD(DL(mul))+SG(ps[index]);
5156: MKNM_ind_pair(pair,mul,index,sugar);
5157: red = ndv_mul_nm_symbolic(mul,ps[index]);
5158: add_pbucket_symbolic(bucket,nd_remove_head(red));
5159: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
5160: }
5161: col++;
5162: }
5163: if ( rp0 ) NEXT(rp) = 0;
5164: NEXT(s) = 0;
5165: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
5166: for ( i = 0, p = s0v, s = s0; i < col;
5167: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
5168: *s0vect = s0v;
5169: *r = rp0;
5170: return col;
1.65 noro 5171: }
5172:
1.69 noro 5173: NODE nd_f4(int m)
5174: {
1.157 noro 5175: int i,nh,stat,index;
5176: NODE r,g;
5177: ND_pairs d,l,t;
5178: ND spol,red;
5179: NDV nf,redv;
5180: NM s0,s;
5181: NODE rp0,srp0,nflist;
5182: int nsp,nred,col,rank,len,k,j,a;
5183: UINT c;
5184: UINT **spmat;
5185: UINT *s0vect,*svect,*p,*v;
5186: int *colstat;
5187: IndArray *imat;
5188: int *rhead;
5189: int spcol,sprow;
5190: int sugar;
5191: PGeoBucket bucket;
5192: struct oEGT eg0,eg1,eg_f4;
1.69 noro 5193:
1.103 noro 5194: #if 0
1.157 noro 5195: ndv_alloc = 0;
1.103 noro 5196: #endif
1.157 noro 5197: g = 0; d = 0;
5198: for ( i = 0; i < nd_psn; i++ ) {
5199: d = update_pairs(d,g,i);
5200: g = update_base(g,i);
5201: }
5202: while ( d ) {
5203: get_eg(&eg0);
5204: l = nd_minsugarp(d,&d);
5205: sugar = SG(l);
5206: bucket = create_pbucket();
5207: stat = nd_sp_f4(m,0,l,bucket);
5208: if ( !stat ) {
5209: for ( t = l; NEXT(t); t = NEXT(t) );
5210: NEXT(t) = d; d = l;
5211: d = nd_reconstruct(0,d);
5212: continue;
5213: }
5214: if ( bucket->m < 0 ) continue;
5215: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
5216: if ( !col ) {
5217: for ( t = l; NEXT(t); t = NEXT(t) );
5218: NEXT(t) = d; d = l;
5219: d = nd_reconstruct(0,d);
5220: continue;
5221: }
5222: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
5223: if ( DP_Print )
5224: fprintf(asir_out,"sugar=%d,symb=%fsec,",
5225: sugar,eg_f4.exectime+eg_f4.gctime);
5226: if ( 1 )
5227: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,0);
5228: else
5229: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0,0);
5230: /* adding new bases */
5231: for ( r = nflist; r; r = NEXT(r) ) {
5232: nf = (NDV)BDY(r);
5233: ndv_removecont(m,nf);
5234: if ( !m && nd_nalg ) {
5235: ND nf1;
5236:
5237: nf1 = ndvtond(m,nf);
5238: nd_monic(0,&nf1);
5239: nd_removecont(m,nf1);
5240: nf = ndtondv(m,nf1);
5241: }
5242: nh = ndv_newps(m,nf,0);
5243: d = update_pairs(d,g,nh);
5244: g = update_base(g,nh);
5245: }
5246: }
5247: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.103 noro 5248: #if 0
1.157 noro 5249: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.103 noro 5250: #endif
1.157 noro 5251: return g;
1.69 noro 5252: }
1.74 noro 5253:
1.133 noro 5254: NODE nd_f4_trace(int m)
5255: {
1.157 noro 5256: int i,nh,stat,index;
5257: NODE r,g;
5258: ND_pairs d,l,l0,t;
5259: ND spol,red;
5260: NDV nf,redv,nfqv,nfv;
5261: NM s0,s;
5262: NODE rp0,srp0,nflist;
5263: int nsp,nred,col,rank,len,k,j,a;
5264: UINT c;
5265: UINT **spmat;
5266: UINT *s0vect,*svect,*p,*v;
5267: int *colstat;
5268: IndArray *imat;
5269: int *rhead;
5270: int spcol,sprow;
5271: int sugar;
5272: PGeoBucket bucket;
5273: struct oEGT eg0,eg1,eg_f4;
5274:
5275: g = 0; d = 0;
5276: for ( i = 0; i < nd_psn; i++ ) {
5277: d = update_pairs(d,g,i);
5278: g = update_base(g,i);
5279: }
5280: while ( d ) {
5281: get_eg(&eg0);
5282: l = nd_minsugarp(d,&d);
5283: sugar = SG(l);
5284: bucket = create_pbucket();
5285: stat = nd_sp_f4(m,0,l,bucket);
5286: if ( !stat ) {
5287: for ( t = l; NEXT(t); t = NEXT(t) );
5288: NEXT(t) = d; d = l;
5289: d = nd_reconstruct(1,d);
5290: continue;
5291: }
5292: if ( bucket->m < 0 ) continue;
5293: col = nd_symbolic_preproc(bucket,0,&s0vect,&rp0);
5294: if ( !col ) {
5295: for ( t = l; NEXT(t); t = NEXT(t) );
5296: NEXT(t) = d; d = l;
5297: d = nd_reconstruct(1,d);
5298: continue;
5299: }
5300: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
5301: if ( DP_Print )
5302: fprintf(asir_out,"sugar=%d,symb=%fsec,",
5303: sugar,eg_f4.exectime+eg_f4.gctime);
5304: nflist = nd_f4_red(m,l,0,s0vect,col,rp0,&l0);
5305: if ( !l0 ) continue;
5306: l = l0;
5307:
5308: /* over Q */
5309: bucket = create_pbucket();
5310: stat = nd_sp_f4(0,1,l,bucket);
5311: if ( !stat ) {
5312: for ( t = l; NEXT(t); t = NEXT(t) );
5313: NEXT(t) = d; d = l;
5314: d = nd_reconstruct(1,d);
5315: continue;
5316: }
5317: if ( bucket->m < 0 ) continue;
5318: col = nd_symbolic_preproc(bucket,1,&s0vect,&rp0);
5319: if ( !col ) {
5320: for ( t = l; NEXT(t); t = NEXT(t) );
5321: NEXT(t) = d; d = l;
5322: d = nd_reconstruct(1,d);
5323: continue;
5324: }
5325: nflist = nd_f4_red(0,l,1,s0vect,col,rp0,0);
5326: /* adding new bases */
5327: for ( r = nflist; r; r = NEXT(r) ) {
5328: nfqv = (NDV)BDY(r);
5329: ndv_removecont(0,nfqv);
5330: if ( !rem(NM(HCQ(nfqv)),m) ) return 0;
5331: if ( nd_nalg ) {
5332: ND nf1;
5333:
5334: nf1 = ndvtond(m,nfqv);
5335: nd_monic(0,&nf1);
5336: nd_removecont(0,nf1);
5337: nfqv = ndtondv(0,nf1); nd_free(nf1);
5338: }
5339: nfv = ndv_dup(0,nfqv);
5340: ndv_mod(m,nfv);
5341: ndv_removecont(m,nfv);
5342: nh = ndv_newps(0,nfv,nfqv);
5343: d = update_pairs(d,g,nh);
5344: g = update_base(g,nh);
5345: }
5346: }
5347: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps_trace[(int)BDY(r)];
1.133 noro 5348: #if 0
1.157 noro 5349: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.133 noro 5350: #endif
1.157 noro 5351: return g;
1.133 noro 5352: }
5353:
5354: NODE nd_f4_red(int m,ND_pairs sp0,int trace,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.63 noro 5355: {
1.157 noro 5356: IndArray *imat;
5357: int nsp,nred,i;
5358: int *rhead;
5359: NODE r0,rp;
5360: ND_pairs sp;
5361: NM_ind_pair *rvect;
5362:
5363: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5364: nred = length(rp0);
5365: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
5366: rhead = (int *)ALLOCA(col*sizeof(int));
5367: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5368:
5369: /* construction of index arrays */
5370: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
5371: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
5372: rvect[i] = (NM_ind_pair)BDY(rp);
5373: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
5374: rhead[imat[i]->head] = 1;
5375: }
5376: if ( m )
5377: r0 = nd_f4_red_main(m,sp0,nsp,s0vect,col,rvect,rhead,imat,nred,nz);
5378: else
5379: r0 = nd_f4_red_q_main(sp0,nsp,trace,s0vect,col,rvect,rhead,imat,nred);
5380: return r0;
1.106 noro 5381: }
1.74 noro 5382:
1.106 noro 5383: NODE nd_f4_red_main(int m,ND_pairs sp0,int nsp,UINT *s0vect,int col,
1.133 noro 5384: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred,ND_pairs *nz)
1.106 noro 5385: {
1.157 noro 5386: int spcol,sprow,a;
5387: int i,j,k,l,rank;
5388: NODE r0,r;
5389: ND_pairs sp;
5390: ND spol;
5391: int **spmat;
5392: UINT *svect,*v;
5393: int *colstat;
5394: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
5395: int maxrs;
5396: int *spsugar;
5397: ND_pairs *spactive;
5398:
5399: spcol = col-nred;
5400: get_eg(&eg0);
5401: /* elimination (1st step) */
5402: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
5403: svect = (UINT *)ALLOCA(col*sizeof(UINT));
5404: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
5405: spactive = !nz?0:(ND_pairs *)ALLOCA(nsp*sizeof(ND_pairs));
5406: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
5407: nd_sp(m,0,sp,&spol);
5408: if ( !spol ) continue;
5409: nd_to_vect(m,s0vect,col,spol,svect);
5410: if ( m == -1 )
5411: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
5412: else
5413: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
5414: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5415: if ( i < col ) {
5416: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
5417: for ( j = k = 0; j < col; j++ )
5418: if ( !rhead[j] ) v[k++] = svect[j];
5419: spsugar[sprow] = MAX(maxrs,SG(spol));
5420: if ( nz )
5421: spactive[sprow] = sp;
5422: sprow++;
5423: }
5424: nd_free(spol);
5425: }
5426: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
5427: if ( DP_Print ) {
5428: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
5429: fflush(asir_out);
5430: }
5431: /* free index arrays */
5432: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
5433:
5434: /* elimination (2nd step) */
5435: colstat = (int *)ALLOCA(spcol*sizeof(int));
5436: if ( m == -1 )
5437: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
5438: else
5439: rank = nd_gauss_elim_mod(spmat,spsugar,spactive,sprow,spcol,m,colstat);
5440: r0 = 0;
5441: for ( i = 0; i < rank; i++ ) {
5442: NEXTNODE(r0,r); BDY(r) =
5443: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5444: SG((NDV)BDY(r)) = spsugar[i];
5445: GC_free(spmat[i]);
5446: }
5447: if ( r0 ) NEXT(r) = 0;
5448:
5449: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
5450: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
5451: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
5452: if ( DP_Print ) {
5453: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
5454: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
5455: nsp,nred,sprow,spcol,rank);
5456: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
5457: }
5458: if ( nz ) {
5459: for ( i = 0; i < rank-1; i++ ) NEXT(spactive[i]) = spactive[i+1];
5460: if ( rank > 0 ) {
5461: NEXT(spactive[rank-1]) = 0;
5462: *nz = spactive[0];
5463: } else
5464: *nz = 0;
5465: }
5466: return r0;
1.74 noro 5467: }
5468:
1.133 noro 5469: #if 1
5470: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,int trace,UINT *s0vect,int col,
1.107 noro 5471: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
5472: {
1.157 noro 5473: int spcol,sprow,a;
5474: int i,j,k,l,rank;
5475: NODE r0,r;
5476: ND_pairs sp;
5477: ND spol;
5478: Q **spmat;
5479: Q *svect,*v;
5480: int *colstat;
5481: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
5482: int maxrs;
5483: int *spsugar;
5484: pointer *w;
5485:
5486: spcol = col-nred;
5487: get_eg(&eg0);
5488: /* elimination (1st step) */
5489: spmat = (Q **)ALLOCA(nsp*sizeof(Q *));
5490: svect = (Q *)ALLOCA(col*sizeof(Q));
5491: spsugar = (int *)ALLOCA(nsp*sizeof(Q));
5492: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
5493: nd_sp(0,trace,sp,&spol);
5494: if ( !spol ) continue;
5495: nd_to_vect_q(s0vect,col,spol,svect);
5496: maxrs = ndv_reduce_vect_q(svect,trace,col,imat,rvect,nred);
5497: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5498: if ( i < col ) {
5499: spmat[sprow] = v = (Q *)MALLOC(spcol*sizeof(Q));
5500: for ( j = k = 0; j < col; j++ )
5501: if ( !rhead[j] ) v[k++] = svect[j];
5502: spsugar[sprow] = MAX(maxrs,SG(spol));
5503: sprow++;
5504: }
5505: /* nd_free(spol); */
5506: }
5507: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
5508: if ( DP_Print ) {
5509: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
5510: fflush(asir_out);
5511: }
5512: /* free index arrays */
5513: /* for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c); */
5514:
5515: /* elimination (2nd step) */
5516: colstat = (int *)ALLOCA(spcol*sizeof(int));
5517: rank = nd_gauss_elim_q(spmat,spsugar,sprow,spcol,colstat);
5518: w = (pointer *)ALLOCA(rank*sizeof(pointer));
5519: for ( i = 0; i < rank; i++ ) {
5520: w[rank-i-1] = (pointer)vect_to_ndv_q(spmat[i],spcol,col,rhead,s0vect);
5521: SG((NDV)w[rank-i-1]) = spsugar[i];
5522: /* GC_free(spmat[i]); */
5523: }
1.138 noro 5524: #if 0
1.157 noro 5525: qsort(w,rank,sizeof(NDV),
5526: (int (*)(const void *,const void *))ndv_compare);
1.137 noro 5527: #endif
1.157 noro 5528: r0 = 0;
5529: for ( i = 0; i < rank; i++ ) {
5530: NEXTNODE(r0,r); BDY(r) = w[i];
5531: }
5532: if ( r0 ) NEXT(r) = 0;
5533:
5534: /* for ( ; i < sprow; i++ ) GC_free(spmat[i]); */
5535: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
5536: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
5537: if ( DP_Print ) {
5538: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
5539: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
5540: nsp,nred,sprow,spcol,rank);
5541: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
5542: }
5543: return r0;
1.107 noro 5544: }
1.129 noro 5545: #else
5546: void printm(Q **mat,int row,int col)
5547: {
1.157 noro 5548: int i,j;
5549: printf("[");
5550: for ( i = 0; i < row; i++ ) {
5551: for ( j = 0; j < col; j++ ) {
5552: printexpr(CO,mat[i][j]); printf(" ");
5553: }
5554: printf("]\n");
5555: }
1.129 noro 5556: }
5557:
5558: NODE nd_f4_red_q_main(ND_pairs sp0,int nsp,UINT *s0vect,int col,
5559: NM_ind_pair *rvect,int *rhead,IndArray *imat,int nred)
5560: {
1.157 noro 5561: int row,a;
5562: int i,j,rank;
5563: NODE r0,r;
5564: ND_pairs sp;
5565: ND spol;
5566: Q **mat;
5567: int *colstat;
5568: int *sugar;
5569:
5570: row = nsp+nred;
5571: /* make the matrix */
5572: mat = (Q **)ALLOCA(row*sizeof(Q *));
5573: sugar = (int *)ALLOCA(row*sizeof(int));
5574: for ( row = a = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
5575: nd_sp(0,0,sp,&spol);
5576: if ( !spol ) continue;
5577: mat[row] = (Q *)MALLOC(col*sizeof(Q));
5578: nd_to_vect_q(s0vect,col,spol,mat[row]);
5579: sugar[row] = SG(spol);
5580: row++;
5581: }
5582: for ( i = 0; i < nred; i++, row++ ) {
5583: mat[row] = nm_ind_pair_to_vect(0,s0vect,col,rvect[i]);
5584: sugar[row] = rvect[i]->sugar;
5585: }
5586: /* elimination */
5587: colstat = (int *)ALLOCA(col*sizeof(int));
5588: rank = nd_gauss_elim_q(mat,sugar,row,col,colstat);
5589: r0 = 0;
5590: for ( i = 0; i < rank; i++ ) {
5591: for ( j = 0; j < col; j++ ) if ( mat[i][j] ) break;
5592: if ( j == col ) error("nd_f4_red_q_main : cannot happen");
5593: if ( rhead[j] ) continue;
5594: NEXTNODE(r0,r); BDY(r) =
5595: (pointer)plain_vect_to_ndv_q(mat[i],col,s0vect);
5596: SG((NDV)BDY(r)) = sugar[i];
5597: }
5598: if ( r0 ) NEXT(r) = 0;
5599: printf("\n");
5600: return r0;
1.129 noro 5601: }
5602: #endif
1.107 noro 5603:
1.74 noro 5604: FILE *nd_write,*nd_read;
5605:
5606: void nd_send_int(int a) {
1.157 noro 5607: write_int(nd_write,&a);
1.74 noro 5608: }
5609:
5610: void nd_send_intarray(int *p,int len) {
1.157 noro 5611: write_intarray(nd_write,p,len);
1.74 noro 5612: }
5613:
5614: int nd_recv_int() {
1.157 noro 5615: int a;
1.74 noro 5616:
1.157 noro 5617: read_int(nd_read,&a);
5618: return a;
1.74 noro 5619: }
5620:
5621: void nd_recv_intarray(int *p,int len) {
1.157 noro 5622: read_intarray(nd_read,p,len);
1.74 noro 5623: }
5624:
5625: void nd_send_ndv(NDV p) {
1.157 noro 5626: int len,i;
5627: NMV m;
1.74 noro 5628:
1.157 noro 5629: if ( !p ) nd_send_int(0);
5630: else {
5631: len = LEN(p);
5632: nd_send_int(len);
5633: m = BDY(p);
5634: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5635: nd_send_int(CM(m));
5636: nd_send_intarray(DL(m),nd_wpd);
5637: }
5638: }
1.74 noro 5639: }
5640:
5641: void nd_send_nd(ND p) {
1.157 noro 5642: int len,i;
5643: NM m;
1.74 noro 5644:
1.157 noro 5645: if ( !p ) nd_send_int(0);
5646: else {
5647: len = LEN(p);
5648: nd_send_int(len);
5649: m = BDY(p);
5650: for ( i = 0; i < len; i++, m = NEXT(m) ) {
5651: nd_send_int(CM(m));
5652: nd_send_intarray(DL(m),nd_wpd);
5653: }
5654: }
1.74 noro 5655: }
1.65 noro 5656:
1.74 noro 5657: NDV nd_recv_ndv()
5658: {
1.157 noro 5659: int len,i;
5660: NMV m,m0;
5661: NDV r;
5662:
5663: len = nd_recv_int();
5664: if ( !len ) return 0;
5665: else {
5666: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
1.103 noro 5667: #if 0
1.157 noro 5668: ndv_alloc += len*nmv_adv;
1.103 noro 5669: #endif
1.157 noro 5670: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
5671: CM(m) = nd_recv_int();
5672: nd_recv_intarray(DL(m),nd_wpd);
5673: }
5674: MKNDV(nd_nvar,m0,len,r);
5675: return r;
5676: }
1.74 noro 5677: }
1.65 noro 5678:
1.74 noro 5679: int ox_exec_f4_red(Q proc)
5680: {
1.157 noro 5681: Obj obj;
5682: STRING fname;
5683: NODE arg;
5684: int s;
5685: extern int ox_need_conv,ox_file_io;
5686:
5687: MKSTR(fname,"nd_exec_f4_red");
5688: arg = mknode(2,proc,fname);
5689: Pox_cmo_rpc(arg,&obj);
5690: s = get_ox_server_id(QTOS(proc));
5691: nd_write = iofp[s].out;
5692: nd_read = iofp[s].in;
5693: ox_need_conv = ox_file_io = 0;
5694: return s;
1.74 noro 5695: }
5696:
1.133 noro 5697: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0,ND_pairs *nz)
1.74 noro 5698: {
1.157 noro 5699: int nsp,nred;
5700: int i,rank,s;
5701: NODE rp,r0,r;
5702: ND_pairs sp;
5703: NM_ind_pair pair;
5704: NMV nmv;
5705: NM nm;
5706: NDV nf;
5707: Obj proc,dmy;
5708:
5709: ox_launch_main(0,0,&proc);
5710: s = ox_exec_f4_red((Q)proc);
5711:
5712: nd_send_int(m);
5713: nd_send_int(nd_nvar);
5714: nd_send_int(nd_bpe);
5715: nd_send_int(nd_wpd);
5716: nd_send_int(nmv_adv);
5717:
5718: saveobj(nd_write,dp_current_spec->obj); fflush(nd_write);
5719:
5720: nd_send_int(nd_psn);
5721: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
5722:
5723: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
5724: nd_send_int(nsp);
5725: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
5726: nd_send_int(sp->i1); nd_send_int(sp->i2);
5727: }
5728:
5729: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
5730:
5731: nred = length(rp0); nd_send_int(nred);
5732: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
5733: pair = (NM_ind_pair)BDY(rp);
5734: nd_send_int(pair->index);
5735: nd_send_intarray(pair->mul->dl,nd_wpd);
5736: }
5737: fflush(nd_write);
5738: rank = nd_recv_int();
5739: fprintf(asir_out,"rank=%d\n",rank);
5740: r0 = 0;
5741: for ( i = 0; i < rank; i++ ) {
5742: nf = nd_recv_ndv();
5743: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
5744: }
5745: Pox_shutdown(mknode(1,proc),&dmy);
5746: return r0;
1.74 noro 5747: }
5748:
5749: /* server side */
5750:
5751: void nd_exec_f4_red_dist()
5752: {
1.157 noro 5753: int m,i,nsp,col,s0size,nred,spcol,j,k;
5754: NM_ind_pair *rp0;
5755: NDV nf;
5756: UINT *s0vect;
5757: IndArray *imat;
5758: int *rhead;
5759: int **spmat;
5760: UINT *svect,*v;
5761: ND_pairs *sp0;
5762: int *colstat;
5763: int a,sprow,rank;
5764: struct order_spec *ord;
5765: Obj ordspec;
5766: ND spol;
5767: int maxrs;
5768: int *spsugar;
5769:
5770: nd_read = iofp[0].in;
5771: nd_write = iofp[0].out;
5772: m = nd_recv_int();
5773: nd_nvar = nd_recv_int();
5774: nd_bpe = nd_recv_int();
5775: nd_wpd = nd_recv_int();
5776: nmv_adv = nd_recv_int();
5777:
5778: loadobj(nd_read,&ordspec);
5779: create_order_spec(0,ordspec,&ord);
5780: nd_init_ord(ord);
5781: nd_setup_parameters(nd_nvar,0);
5782:
5783: nd_psn = nd_recv_int();
5784: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
5785: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
5786: for ( i = 0; i < nd_psn; i++ ) {
5787: nd_ps[i] = nd_recv_ndv();
5788: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
5789: }
5790:
5791: nsp = nd_recv_int();
5792: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
5793: for ( i = 0; i < nsp; i++ ) {
5794: NEWND_pairs(sp0[i]);
5795: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
5796: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
5797: }
5798:
5799: col = nd_recv_int();
5800: s0size = col*nd_wpd;
5801: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
5802: nd_recv_intarray(s0vect,s0size);
5803:
5804: nred = nd_recv_int();
5805: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
5806: for ( i = 0; i < nred; i++ ) {
5807: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
5808: rp0[i]->index = nd_recv_int();
5809: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
5810: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
5811: }
5812:
5813: spcol = col-nred;
5814: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
5815: rhead = (int *)MALLOC(col*sizeof(int));
5816: for ( i = 0; i < col; i++ ) rhead[i] = 0;
5817:
5818: /* construction of index arrays */
5819: for ( i = 0; i < nred; i++ ) {
5820: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
5821: rhead[imat[i]->head] = 1;
5822: }
5823:
5824: /* elimination (1st step) */
5825: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
5826: svect = (UINT *)MALLOC(col*sizeof(UINT));
5827: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
5828: for ( a = sprow = 0; a < nsp; a++ ) {
5829: nd_sp(m,0,sp0[a],&spol);
5830: if ( !spol ) continue;
5831: nd_to_vect(m,s0vect,col,spol,svect);
5832: if ( m == -1 )
5833: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
5834: else
5835: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
5836: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
5837: if ( i < col ) {
5838: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
5839: for ( j = k = 0; j < col; j++ )
5840: if ( !rhead[j] ) v[k++] = svect[j];
5841: spsugar[sprow] = MAX(maxrs,SG(spol));
5842: sprow++;
5843: }
5844: nd_free(spol);
5845: }
5846: /* elimination (2nd step) */
5847: colstat = (int *)ALLOCA(spcol*sizeof(int));
5848: if ( m == -1 )
5849: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
5850: else
5851: rank = nd_gauss_elim_mod(spmat,spsugar,0,sprow,spcol,m,colstat);
5852: nd_send_int(rank);
5853: for ( i = 0; i < rank; i++ ) {
5854: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
5855: nd_send_ndv(nf);
5856: }
5857: fflush(nd_write);
1.107 noro 5858: }
5859:
1.113 noro 5860: int nd_gauss_elim_q(Q **mat0,int *sugar,int row,int col,int *colstat)
1.107 noro 5861: {
1.157 noro 5862: int mod,i,j,t,c,rank,rank0,inv;
5863: int *ci,*ri;
5864: Q dn;
5865: MAT m,nm;
5866: int **wmat;
5867:
5868: /* XXX */
5869: mod = 99999989;
5870: wmat = (int **)ALLOCA(row*sizeof(int *));
5871: for ( i = 0; i < row; i++ ) {
5872: wmat[i] = (int *)ALLOCA(col*sizeof(int));
5873: for ( j = 0; j < col; j++ ) {
5874: if ( mat0[i][j] ) {
5875: t = rem(NM(mat0[i][j]),mod);
5876: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5877: wmat[i][j] = t;
5878: } else
5879: wmat[i][j] = 0;
5880: }
5881: }
5882: rank0 = nd_gauss_elim_mod(wmat,sugar,0,row,col,mod,colstat);
5883: NEWMAT(m); m->row = row; m->col = col; m->body = (pointer **)mat0;
5884: rank = generic_gauss_elim(m,&nm,&dn,&ri,&ci);
5885: if ( rank != rank0 )
5886: error("afo");
5887: for ( i = 0; i < row; i++ )
5888: for ( j = 0; j < col; j++ )
5889: mat0[i][j] = 0;
5890: c = col-rank;
5891: for ( i = 0; i < rank; i++ ) {
5892: mat0[i][ri[i]] = dn;
5893: for ( j = 0; j < c; j++ )
5894: mat0[i][ci[j]] = (Q)BDY(nm)[i][j];
5895: }
5896: inv = invm(rem(NM(dn),mod),mod);
5897: if ( SGN(dn) < 0 ) inv = mod-inv;
5898: for ( i = 0; i < row; i++ )
5899: for ( j = 0; j < col; j++ ) {
5900: if ( mat0[i][j] ) {
5901: t = rem(NM(mat0[i][j]),mod);
5902: if ( SGN(mat0[i][j]) < 0 ) t = mod-t;
5903: } else
5904: t = 0;
5905: c = dmar(t,inv,0,mod);
5906: if ( wmat[i][j] != c )
5907: error("afo");
5908: }
5909: return rank;
1.76 noro 5910: }
5911:
1.133 noro 5912: int nd_gauss_elim_mod(int **mat0,int *sugar,ND_pairs *spactive,int row,int col,int md,int *colstat)
1.76 noro 5913: {
1.157 noro 5914: int i,j,k,l,inv,a,rank,s;
5915: unsigned int *t,*pivot,*pk;
5916: unsigned int **mat;
5917: ND_pairs pair;
5918:
5919: mat = (unsigned int **)mat0;
5920: for ( rank = 0, j = 0; j < col; j++ ) {
5921: for ( i = rank; i < row; i++ )
5922: mat[i][j] %= md;
5923: for ( i = rank; i < row; i++ )
5924: if ( mat[i][j] )
5925: break;
5926: if ( i == row ) {
5927: colstat[j] = 0;
5928: continue;
5929: } else
5930: colstat[j] = 1;
5931: if ( i != rank ) {
5932: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
5933: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
5934: if ( spactive ) {
5935: pair = spactive[i]; spactive[i] = spactive[rank];
5936: spactive[rank] = pair;
5937: }
5938: }
5939: pivot = mat[rank];
5940: s = sugar[rank];
5941: inv = invm(pivot[j],md);
5942: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
5943: if ( *pk ) {
5944: if ( *pk >= (unsigned int)md )
5945: *pk %= md;
5946: DMAR(*pk,inv,0,md,*pk)
5947: }
5948: for ( i = rank+1; i < row; i++ ) {
5949: t = mat[i];
5950: if ( a = t[j] ) {
5951: sugar[i] = MAX(sugar[i],s);
5952: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5953: }
5954: }
5955: rank++;
5956: }
5957: for ( j = col-1, l = rank-1; j >= 0; j-- )
5958: if ( colstat[j] ) {
5959: pivot = mat[l];
5960: s = sugar[l];
5961: for ( i = 0; i < l; i++ ) {
5962: t = mat[i];
5963: t[j] %= md;
5964: if ( a = t[j] ) {
5965: sugar[i] = MAX(sugar[i],s);
5966: red_by_vect(md,t+j,pivot+j,md-a,col-j);
5967: }
5968: }
5969: l--;
5970: }
5971: for ( j = 0, l = 0; l < rank; j++ )
5972: if ( colstat[j] ) {
5973: t = mat[l];
5974: for ( k = j; k < col; k++ )
5975: if ( t[k] >= (unsigned int)md )
5976: t[k] %= md;
5977: l++;
5978: }
5979: return rank;
1.76 noro 5980: }
5981:
5982: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
5983: {
1.157 noro 5984: int i,j,k,l,inv,a,rank,s;
5985: unsigned int *t,*pivot,*pk;
5986: unsigned int **mat;
5987:
5988: mat = (unsigned int **)mat0;
5989: for ( rank = 0, j = 0; j < col; j++ ) {
5990: for ( i = rank; i < row; i++ )
5991: if ( mat[i][j] )
5992: break;
5993: if ( i == row ) {
5994: colstat[j] = 0;
5995: continue;
5996: } else
5997: colstat[j] = 1;
5998: if ( i != rank ) {
5999: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
6000: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
6001: }
6002: pivot = mat[rank];
6003: s = sugar[rank];
6004: inv = _invsf(pivot[j]);
6005: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
6006: if ( *pk )
6007: *pk = _mulsf(*pk,inv);
6008: for ( i = rank+1; i < row; i++ ) {
6009: t = mat[i];
6010: if ( a = t[j] ) {
6011: sugar[i] = MAX(sugar[i],s);
6012: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
6013: }
6014: }
6015: rank++;
6016: }
6017: for ( j = col-1, l = rank-1; j >= 0; j-- )
6018: if ( colstat[j] ) {
6019: pivot = mat[l];
6020: s = sugar[l];
6021: for ( i = 0; i < l; i++ ) {
6022: t = mat[i];
6023: if ( a = t[j] ) {
6024: sugar[i] = MAX(sugar[i],s);
6025: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
6026: }
6027: }
6028: l--;
6029: }
6030: return rank;
1.77 noro 6031: }
6032:
6033: int ndv_ishomo(NDV p)
6034: {
1.157 noro 6035: NMV m;
6036: int len,h;
1.77 noro 6037:
1.157 noro 6038: if ( !p ) return 1;
6039: len = LEN(p);
6040: m = BDY(p);
6041: h = TD(DL(m));
6042: NMV_ADV(m);
6043: for ( len--; len; len--, NMV_ADV(m) )
6044: if ( TD(DL(m)) != h ) return 0;
6045: return 1;
1.77 noro 6046: }
6047:
6048: void ndv_save(NDV p,int index)
6049: {
1.157 noro 6050: FILE *s;
6051: char name[BUFSIZ];
6052: short id;
6053: int nv,sugar,len,n,i,td,e,j;
6054: NMV m;
6055: unsigned int *dl;
6056: int mpos;
6057:
6058: sprintf(name,"%s/%d",Demand,index);
6059: s = fopen(name,"w");
6060: savevl(s,0);
6061: if ( !p ) {
6062: saveobj(s,0);
6063: return;
6064: }
6065: id = O_DP;
6066: nv = NV(p);
6067: sugar = SG(p);
6068: len = LEN(p);
6069: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
6070: write_int(s,&len);
6071:
6072: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
6073: saveobj(s,(Obj)CQ(m));
6074: dl = DL(m);
6075: td = TD(dl);
6076: write_int(s,&td);
6077: for ( j = 0; j < nv; j++ ) {
6078: e = GET_EXP(dl,j);
6079: write_int(s,&e);
6080: }
6081: if ( nd_module ) {
6082: mpos = MPOS(dl); write_int(s,&mpos);
6083: }
6084: }
6085: fclose(s);
1.77 noro 6086: }
6087:
6088: NDV ndv_load(int index)
6089: {
1.157 noro 6090: FILE *s;
6091: char name[BUFSIZ];
6092: short id;
6093: int nv,sugar,len,n,i,td,e,j;
6094: NDV d;
6095: NMV m0,m;
6096: unsigned int *dl;
6097: Obj obj;
6098: int mpos;
6099:
6100: sprintf(name,"%s/%d",Demand,index);
6101: s = fopen(name,"r");
6102: if ( !s ) return 0;
6103:
6104: skipvl(s);
6105: read_short(s,&id);
6106: if ( !id ) return 0;
6107: read_int(s,&nv);
6108: read_int(s,&sugar);
6109: read_int(s,&len);
6110:
6111: m0 = m = MALLOC(len*nmv_adv);
6112: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
6113: loadobj(s,&obj); CQ(m) = (Q)obj;
6114: dl = DL(m);
6115: ndl_zero(dl);
6116: read_int(s,&td); TD(dl) = td;
6117: for ( j = 0; j < nv; j++ ) {
6118: read_int(s,&e);
6119: PUT_EXP(dl,j,e);
6120: }
6121: if ( nd_module ) {
6122: read_int(s,&mpos); MPOS(dl) = mpos;
6123: }
6124: if ( nd_blockmask ) ndl_weight_mask(dl);
6125: }
6126: fclose(s);
6127: MKNDV(nv,m0,len,d);
6128: SG(d) = sugar;
6129: return d;
1.99 noro 6130: }
6131:
1.102 noro 6132: void nd_det(int mod,MAT f,P *rp)
1.99 noro 6133: {
1.157 noro 6134: VL fv,tv;
6135: int n,i,j,max,e,nvar,sgn,k0,l0,len0,len,k,l,a;
6136: pointer **m;
6137: Q mone;
6138: P **w;
6139: P mp,r;
6140: NDV **dm;
6141: NDV *t,*mi,*mj;
6142: NDV d,s,mij,mjj;
6143: ND u;
6144: NMV nmv;
6145: UINT *bound;
6146: PGeoBucket bucket;
6147: struct order_spec *ord;
6148: Q dq,dt,ds;
6149: N gn,qn,dn0,nm,dn;
6150:
6151: create_order_spec(0,0,&ord);
6152: nd_init_ord(ord);
6153: get_vars((Obj)f,&fv);
6154: if ( f->row != f->col )
6155: error("nd_det : non-square matrix");
6156: n = f->row;
6157: m = f->body;
6158: for ( nvar = 0, tv = fv; tv; tv = NEXT(tv), nvar++ );
6159:
6160: if ( !nvar ) {
6161: if ( !mod )
6162: detp(CO,(P **)m,n,rp);
6163: else {
6164: w = (P **)almat_pointer(n,n);
6165: for ( i = 0; i < n; i++ )
6166: for ( j = 0; j < n; j++ )
6167: ptomp(mod,(P)m[i][j],&w[i][j]);
6168: detmp(CO,mod,w,n,&mp);
6169: mptop(mp,rp);
6170: }
6171: return;
6172: }
6173:
6174: if ( !mod ) {
6175: w = (P **)almat_pointer(n,n);
6176: dq = ONE;
6177: for ( i = 0; i < n; i++ ) {
6178: dn0 = ONEN;
6179: for ( j = 0; j < n; j++ ) {
6180: if ( !m[i][j] ) continue;
6181: lgp(m[i][j],&nm,&dn);
6182: gcdn(dn0,dn,&gn); divsn(dn0,gn,&qn); muln(qn,dn,&dn0);
6183: }
6184: if ( !UNIN(dn0) ) {
6185: NTOQ(dn0,1,ds);
6186: for ( j = 0; j < n; j++ )
6187: mulp(CO,(P)m[i][j],(P)ds,&w[i][j]);
6188: mulq(dq,ds,&dt); dq = dt;
6189: } else
6190: for ( j = 0; j < n; j++ )
6191: w[i][j] = (P)m[i][j];
6192: }
6193: m = (pointer **)w;
6194: }
6195:
6196: for ( i = 0, max = 0; i < n; i++ )
6197: for ( j = 0; j < n; j++ )
6198: for ( tv = fv; tv; tv = NEXT(tv) ) {
6199: e = getdeg(tv->v,(P)m[i][j]);
6200: max = MAX(e,max);
6201: }
6202: nd_setup_parameters(nvar,max);
6203: dm = (NDV **)almat_pointer(n,n);
6204: for ( i = 0, max = 0; i < n; i++ )
6205: for ( j = 0; j < n; j++ ) {
6206: dm[i][j] = ptondv(CO,fv,m[i][j]);
6207: if ( mod ) ndv_mod(mod,dm[i][j]);
6208: if ( dm[i][j] && !LEN(dm[i][j]) ) dm[i][j] = 0;
6209: }
6210: d = ptondv(CO,fv,(P)ONE);
6211: if ( mod ) ndv_mod(mod,d);
6212: chsgnq(ONE,&mone);
6213: for ( j = 0, sgn = 1; j < n; j++ ) {
6214: if ( DP_Print ) fprintf(stderr,".",j);
6215: for ( i = j; i < n && !dm[i][j]; i++ );
6216: if ( i == n ) {
6217: *rp = 0;
6218: return;
6219: }
6220: k0 = i; l0 = j; len0 = LEN(dm[k0][l0]);
6221: for ( k = j; k < n; k++ )
6222: for ( l = j; l < n; l++ )
6223: if ( dm[k][l] && LEN(dm[k][l]) < len0 ) {
6224: k0 = k; l0 = l; len0 = LEN(dm[k][l]);
6225: }
6226: if ( k0 != j ) {
6227: t = dm[j]; dm[j] = dm[k0]; dm[k0] = t;
6228: sgn = -sgn;
6229: }
6230: if ( l0 != j ) {
6231: for ( k = j; k < n; k++ ) {
6232: s = dm[k][j]; dm[k][j] = dm[k][l0]; dm[k][l0] = s;
6233: }
6234: sgn = -sgn;
6235: }
6236: bound = nd_det_compute_bound(dm,n,j);
6237: for ( k = 0; k < nd_nvar; k++ )
6238: if ( bound[k]*2 > nd_mask0 ) break;
6239: if ( k < nd_nvar )
6240: nd_det_reconstruct(dm,n,j,d);
6241:
6242: for ( i = j+1, mj = dm[j], mjj = mj[j]; i < n; i++ ) {
6243: /* if ( DP_Print ) fprintf(stderr," i=%d\n ",i); */
6244: mi = dm[i]; mij = mi[j];
6245: if ( mod )
6246: ndv_mul_c(mod,mij,mod-1);
6247: else
6248: ndv_mul_c_q(mij,mone);
6249: for ( k = j+1; k < n; k++ ) {
6250: /* if ( DP_Print ) fprintf(stderr,"k=%d ",k); */
6251: bucket = create_pbucket();
6252: if ( mi[k] ) {
6253: nmv = BDY(mjj); len = LEN(mjj);
6254: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
6255: u = ndv_mul_nmv_trunc(mod,nmv,mi[k],DL(BDY(d)));
6256: add_pbucket(mod,bucket,u);
6257: }
6258: }
6259: if ( mj[k] && mij ) {
6260: nmv = BDY(mij); len = LEN(mij);
6261: for ( a = 0; a < len; a++, NMV_ADV(nmv) ) {
6262: u = ndv_mul_nmv_trunc(mod,nmv,mj[k],DL(BDY(d)));
6263: add_pbucket(mod,bucket,u);
6264: }
6265: }
6266: u = nd_quo(mod,bucket,d);
6267: mi[k] = ndtondv(mod,u);
6268: }
6269: /* if ( DP_Print ) fprintf(stderr,"\n",k); */
6270: }
6271: d = mjj;
6272: }
6273: if ( DP_Print ) fprintf(stderr,"\n",k);
6274: if ( sgn < 0 )
6275: if ( mod )
6276: ndv_mul_c(mod,d,mod-1);
6277: else
6278: ndv_mul_c_q(d,mone);
6279: r = ndvtop(mod,CO,fv,d);
6280: if ( !mod && !UNIQ(dq) )
6281: divsp(CO,r,(P)dq,rp);
6282: else
6283: *rp = r;
1.99 noro 6284: }
6285:
1.102 noro 6286: ND ndv_mul_nmv_trunc(int mod,NMV m0,NDV p,UINT *d)
1.99 noro 6287: {
1.157 noro 6288: NM mr,mr0;
6289: NM tnm;
6290: NMV m;
6291: UINT *d0,*dt,*dm;
6292: int c,n,td,i,c1,c2,len;
6293: Q q;
6294: ND r;
6295:
6296: if ( !p ) return 0;
6297: else {
6298: n = NV(p); m = BDY(p); len = LEN(p);
6299: d0 = DL(m0);
6300: td = TD(d);
6301: mr0 = 0;
6302: NEWNM(tnm);
6303: if ( mod ) {
6304: c = CM(m0);
6305: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
6306: ndl_add(DL(m),d0,DL(tnm));
6307: if ( ndl_reducible(DL(tnm),d) ) {
6308: NEXTNM(mr0,mr);
6309: c1 = CM(m); DMAR(c1,c,0,mod,c2); CM(mr) = c2;
6310: ndl_copy(DL(tnm),DL(mr));
6311: }
6312: }
6313: } else {
6314: q = CQ(m0);
6315: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
6316: ndl_add(DL(m),d0,DL(tnm));
6317: if ( ndl_reducible(DL(tnm),d) ) {
6318: NEXTNM(mr0,mr);
6319: mulq(CQ(m),q,&CQ(mr));
6320: ndl_copy(DL(tnm),DL(mr));
6321: }
6322: }
6323: }
6324: if ( !mr0 )
6325: return 0;
6326: else {
6327: NEXT(mr) = 0;
6328: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
6329: MKND(NV(p),mr0,len,r);
6330: SG(r) = SG(p) + TD(d0);
6331: return r;
6332: }
6333: }
1.114 noro 6334: }
6335:
6336: void nd_det_reconstruct(NDV **dm,int n,int j,NDV d)
6337: {
1.157 noro 6338: int i,obpe,oadv,h,k,l;
6339: static NM prev_nm_free_list;
6340: EPOS oepos;
6341:
6342: obpe = nd_bpe;
6343: oadv = nmv_adv;
6344: oepos = nd_epos;
6345: if ( obpe < 2 ) nd_bpe = 2;
6346: else if ( obpe < 3 ) nd_bpe = 3;
6347: else if ( obpe < 4 ) nd_bpe = 4;
6348: else if ( obpe < 5 ) nd_bpe = 5;
6349: else if ( obpe < 6 ) nd_bpe = 6;
6350: else if ( obpe < 8 ) nd_bpe = 8;
6351: else if ( obpe < 10 ) nd_bpe = 10;
6352: else if ( obpe < 16 ) nd_bpe = 16;
6353: else if ( obpe < 32 ) nd_bpe = 32;
6354: else error("nd_det_reconstruct : exponent too large");
6355:
6356: nd_setup_parameters(nd_nvar,0);
6357: prev_nm_free_list = _nm_free_list;
6358: _nm_free_list = 0;
6359: for ( k = j; k < n; k++ )
6360: for (l = j; l < n; l++ )
6361: ndv_realloc(dm[k][l],obpe,oadv,oepos);
6362: ndv_realloc(d,obpe,oadv,oepos);
6363: prev_nm_free_list = 0;
1.114 noro 6364: #if 0
1.157 noro 6365: GC_gcollect();
1.114 noro 6366: #endif
6367: }
6368:
1.153 noro 6369: /* returns a UINT array containing degree bounds */
6370:
1.114 noro 6371: UINT *nd_det_compute_bound(NDV **dm,int n,int j)
6372: {
1.157 noro 6373: UINT *d0,*d1,*d,*t,*r;
6374: int k,l,i;
1.114 noro 6375:
1.157 noro 6376: d0 = (UINT *)MALLOC(nd_nvar*sizeof(UINT));
6377: for ( k = 0; k < nd_nvar; k++ ) d0[k] = 0;
6378: for ( k = j; k < n; k++ )
6379: for ( l = j; l < n; l++ )
6380: if ( dm[k][l] ) {
6381: d = ndv_compute_bound(dm[k][l]);
6382: for ( i = 0; i < nd_nvar; i++ )
6383: d0[i] = MAX(d0[i],d[i]);
6384: }
6385: return d0;
1.117 noro 6386: }
6387:
6388: DL nd_separate_d(UINT *d,UINT *trans)
6389: {
1.157 noro 6390: int n,td,i,e,j;
6391: DL a;
1.117 noro 6392:
1.157 noro 6393: ndl_zero(trans);
6394: td = 0;
6395: for ( i = 0; i < nd_ntrans; i++ ) {
6396: e = GET_EXP(d,i);
6397: PUT_EXP(trans,i,e);
6398: td += MUL_WEIGHT(e,i);
6399: }
6400: if ( nd_ntrans+nd_nalg < nd_nvar ) {
6401: /* homogenized */
6402: i = nd_nvar-1;
6403: e = GET_EXP(d,i);
6404: PUT_EXP(trans,i,e);
6405: td += MUL_WEIGHT(e,i);
6406: }
6407: TD(trans) = td;
6408: if ( nd_blockmask) ndl_weight_mask(trans);
6409: NEWDL(a,nd_nalg);
6410: td = 0;
6411: for ( i = 0; i < nd_nalg; i++ ) {
6412: j = nd_ntrans+i;
6413: e = GET_EXP(d,j);
6414: a->d[i] = e;
6415: td += e;
6416: }
6417: a->td = td;
6418: return a;
1.117 noro 6419: }
6420:
1.118 noro 6421: int nd_monic(int mod,ND *p)
1.117 noro 6422: {
1.157 noro 6423: UINT *trans,*t;
6424: DL alg;
6425: MP mp0,mp;
6426: NM m,m0,m1,ma0,ma,mb,mr0,mr;
6427: ND r;
6428: DL dl;
6429: DP nm;
6430: NDV ndv;
6431: DAlg inv,cd;
6432: ND s,c;
6433: Q l,mul;
6434: N ln;
6435: int n,ntrans,i,e,td,is_lc,len;
6436: NumberField nf;
6437: struct oEGT eg0,eg1;
6438:
6439: if ( !(nf = get_numberfield()) )
6440: error("nd_monic : current_numberfield is not set");
6441:
6442: /* Q coef -> DAlg coef */
6443: NEWNM(ma0); ma = ma0;
6444: m = BDY(*p);
6445: is_lc = 1;
6446: while ( 1 ) {
6447: NEWMP(mp0); mp = mp0;
6448: mp->c = (P)CQ(m);
6449: mp->dl = nd_separate_d(DL(m),DL(ma));
6450: NEWNM(mb);
6451: for ( m = NEXT(m); m; m = NEXT(m) ) {
6452: alg = nd_separate_d(DL(m),DL(mb));
6453: if ( !ndl_equal(DL(ma),DL(mb)) )
6454: break;
6455: NEXTMP(mp0,mp); mp->c = (P)CQ(m); mp->dl = alg;
6456: }
6457: NEXT(mp) = 0;
6458: MKDP(nd_nalg,mp0,nm);
6459: MKDAlg(nm,ONE,cd);
6460: if ( is_lc == 1 ) {
6461: /* if the lc is a rational number, we have nothing to do */
6462: if ( !mp0->dl->td )
6463: return 1;
6464:
6465: get_eg(&eg0);
6466: invdalg(cd,&inv);
6467: get_eg(&eg1); add_eg(&eg_invdalg,&eg0,&eg1);
6468: /* check the validity of inv */
6469: if ( mod && !rem(NM(inv->dn),mod) )
6470: return 0;
6471: CA(ma) = nf->one;
6472: is_lc = 0;
6473: ln = ONEN;
6474: } else {
6475: muldalg(cd,inv,&CA(ma));
6476: lcmn(ln,NM(CA(ma)->dn),&ln);
6477: }
6478: if ( m ) {
6479: NEXT(ma) = mb; ma = mb;
6480: } else {
6481: NEXT(ma) = 0;
6482: break;
6483: }
6484: }
6485: /* l = lcm(denoms) */
6486: NTOQ(ln,1,l);
6487: for ( mr0 = 0, m = ma0; m; m = NEXT(m) ) {
6488: divq(l,CA(m)->dn,&mul);
6489: for ( mp = BDY(CA(m)->nm); mp; mp = NEXT(mp) ) {
6490: NEXTNM(mr0,mr);
6491: mulq((Q)mp->c,mul,&CQ(mr));
6492: dl = mp->dl;
6493: td = TD(DL(m));
6494: ndl_copy(DL(m),DL(mr));
6495: for ( i = 0; i < nd_nalg; i++ ) {
6496: e = dl->d[i];
6497: PUT_EXP(DL(mr),i+nd_ntrans,e);
6498: td += MUL_WEIGHT(e,i+nd_ntrans);
6499: }
6500: TD(DL(mr)) = td;
6501: if ( nd_blockmask) ndl_weight_mask(DL(mr));
6502: if ( nd_module ) MPOS(DL(mr)) = MPOS(DL(m));
6503: }
6504: }
6505: NEXT(mr) = 0;
6506: for ( len = 0, mr = mr0; mr; mr = NEXT(mr), len++ );
6507: MKND(NV(*p),mr0,len,r);
6508: /* XXX */
6509: SG(r) = SG(*p);
6510: nd_free(*p);
6511: *p = r;
6512: return 1;
1.59 noro 6513: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[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