Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.68
1.68 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.67 2003/09/12 14:51:27 noro Exp $ */
1.2 noro 2:
1.1 noro 3: #include "ca.h"
4: #include "inline.h"
1.64 noro 5: #include <time.h>
1.1 noro 6:
7: #if defined(__GNUC__)
8: #define INLINE inline
9: #elif defined(VISUAL)
10: #define INLINE __inline
11: #else
12: #define INLINE
13: #endif
14:
1.61 noro 15: typedef unsigned int UINT;
16:
1.47 noro 17: #define USE_GEOBUCKET 1
1.65 noro 18: #define USE_UNROLL 1
1.28 noro 19:
1.1 noro 20: #define REDTAB_LEN 32003
21:
1.40 noro 22: /* GeoBucket for polynomial addition */
23:
1.1 noro 24: typedef struct oPGeoBucket {
25: int m;
26: struct oND *body[32];
27: } *PGeoBucket;
28:
1.40 noro 29: /* distributed polynomial; linked list rep. */
1.1 noro 30: typedef struct oND {
31: struct oNM *body;
32: int nv;
1.31 noro 33: int len;
1.1 noro 34: int sugar;
35: } *ND;
36:
1.40 noro 37: /* distributed polynomial; array rep. */
1.3 noro 38: typedef struct oNDV {
39: struct oNMV *body;
40: int nv;
1.31 noro 41: int len;
1.3 noro 42: int sugar;
43: } *NDV;
44:
1.40 noro 45: /* monomial; linked list rep. */
1.1 noro 46: typedef struct oNM {
47: struct oNM *next;
1.14 noro 48: union {
49: int m;
50: Q z;
1.61 noro 51: P p;
1.14 noro 52: } c;
1.61 noro 53: UINT dl[1];
1.1 noro 54: } *NM;
55:
1.40 noro 56: /* monomial; array rep. */
1.3 noro 57: typedef struct oNMV {
1.14 noro 58: union {
59: int m;
60: Q z;
1.61 noro 61: P p;
1.14 noro 62: } c;
1.61 noro 63: UINT dl[1];
1.3 noro 64: } *NMV;
65:
1.40 noro 66: /* history of reducer */
1.13 noro 67: typedef struct oRHist {
68: struct oRHist *next;
69: int index;
1.34 noro 70: int sugar;
1.61 noro 71: UINT dl[1];
1.13 noro 72: } *RHist;
73:
1.40 noro 74: /* S-pair list */
1.1 noro 75: typedef struct oND_pairs {
76: struct oND_pairs *next;
77: int i1,i2;
1.34 noro 78: int sugar;
1.61 noro 79: UINT lcm[1];
1.1 noro 80: } *ND_pairs;
81:
1.42 noro 82: /* index and shift count for each exponent */
83: typedef struct oEPOS {
84: int i; /* index */
85: int s; /* shift */
86: } *EPOS;
87:
1.43 noro 88: typedef struct oBlockMask {
89: int n;
90: struct order_pair *order_pair;
1.61 noro 91: UINT **mask;
1.43 noro 92: } *BlockMask;
93:
1.45 noro 94: typedef struct oBaseSet {
95: int len;
96: NDV *ps;
1.61 noro 97: UINT **bound;
1.45 noro 98: } *BaseSet;
99:
1.63 noro 100: typedef struct oNM_ind_pair
101: {
102: NM mul;
103: int index;
104: } *NM_ind_pair;
105:
1.67 noro 106: typedef struct oIndArray
107: {
108: char width;
109: int head;
110: union {
111: unsigned char *c;
112: unsigned short *s;
113: unsigned int *i;
114: } index;
115: } *IndArray;
1.63 noro 116:
1.61 noro 117: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.32 noro 118:
1.42 noro 119: static double nd_scale=2;
1.61 noro 120: static UINT **nd_bound;
1.42 noro 121: static struct order_spec *nd_ord;
122: static EPOS nd_epos;
1.43 noro 123: static BlockMask nd_blockmask;
1.42 noro 124: static int nd_nvar;
125: static int nd_isrlex;
126: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 127: static UINT nd_mask[32];
128: static UINT nd_mask0,nd_mask1;
1.42 noro 129:
130: static NM _nm_free_list;
131: static ND _nd_free_list;
132: static ND_pairs _ndp_free_list;
1.20 noro 133:
134: static NDV *nd_ps;
1.53 noro 135: static NDV *nd_ps_trace;
1.42 noro 136: static RHist *nd_psh;
137: static int nd_psn,nd_pslen;
1.20 noro 138:
1.42 noro 139: static RHist *nd_red;
1.1 noro 140:
1.42 noro 141: static int nd_found,nd_create,nd_notfirst;
142: static int nm_adv;
143: static int nmv_adv;
144: static int nd_dcomp;
1.1 noro 145:
1.61 noro 146: extern VL CO;
1.55 noro 147: extern int Top,Reverse,dp_nelim,do_weyl;
1.58 noro 148: extern int *current_weyl_weight_vector;
1.1 noro 149:
1.40 noro 150: /* fundamental macros */
1.34 noro 151: #define TD(d) (d[0])
1.1 noro 152: #define HDL(d) ((d)->body->dl)
1.34 noro 153: #define HTD(d) (TD(HDL(d)))
1.14 noro 154: #define HCM(d) ((d)->body->c.m)
1.16 noro 155: #define HCQ(d) ((d)->body->c.z)
1.61 noro 156: #define HCP(d) ((d)->body->c.p)
1.14 noro 157: #define CM(a) ((a)->c.m)
1.16 noro 158: #define CQ(a) ((a)->c.z)
1.61 noro 159: #define CP(a) ((a)->c.p)
1.14 noro 160: #define DL(a) ((a)->dl)
161: #define SG(a) ((a)->sugar)
162: #define LEN(a) ((a)->len)
1.33 noro 163: #define LCM(a) ((a)->lcm)
1.42 noro 164: #define GET_EXP(d,a) (((d)[nd_epos[a].i]>>nd_epos[a].s)&nd_mask0)
1.60 noro 165: #define GET_EXP_MASK(d,a,m) ((((d)[nd_epos[a].i]&(m)[nd_epos[a].i])>>nd_epos[a].s)&nd_mask0)
1.42 noro 166: #define PUT_EXP(r,a,e) ((r)[nd_epos[a].i] |= ((e)<<nd_epos[a].s))
1.45 noro 167: #define XOR_EXP(r,a,e) ((r)[nd_epos[a].i] ^= ((e)<<nd_epos[a].s))
1.1 noro 168:
1.61 noro 169: #define GET_EXP_OLD(d,a) (((d)[oepos[a].i]>>oepos[a].s)&omask0)
170: #define PUT_EXP_OLD(r,a,e) ((r)[oepos[a].i] |= ((e)<<oepos[a].s))
171:
1.40 noro 172: /* macros for term comparison */
1.34 noro 173: #define TD_DL_COMPARE(d1,d2)\
1.41 noro 174: (TD(d1)>TD(d2)?1:(TD(d1)<TD(d2)?-1:ndl_lex_compare(d1,d2)))
1.43 noro 175: #if 0
1.34 noro 176: #define DL_COMPARE(d1,d2)\
1.43 noro 177: (nd_dcomp>0?TD_DL_COMPARE(d1,d2)\
178: :(nd_dcomp==0?ndl_lex_compare(d1,d2)\
179: :(nd_blockmask?ndl_block_compare(d1,d2)\
1.45 noro 180: :(*ndl_compare_function)(d1,d2))))
1.43 noro 181: #else
182: #define DL_COMPARE(d1,d2)\
1.45 noro 183: (nd_dcomp>0?TD_DL_COMPARE(d1,d2):(*ndl_compare_function)(d1,d2))
1.43 noro 184: #endif
1.34 noro 185:
1.40 noro 186: /* allocators */
1.15 noro 187: #define NEWRHist(r) \
1.61 noro 188: ((r)=(RHist)MALLOC(sizeof(struct oRHist)+(nd_wpd-1)*sizeof(UINT)))
1.34 noro 189: #define NEWND_pairs(m) \
190: if(!_ndp_free_list)_NDP_alloc();\
191: (m)=_ndp_free_list; _ndp_free_list = NEXT(_ndp_free_list)
192: #define NEWNM(m)\
193: if(!_nm_free_list)_NM_alloc();\
194: (m)=_nm_free_list; _nm_free_list = NEXT(_nm_free_list)
195: #define MKND(n,m,len,d)\
196: if(!_nd_free_list)_ND_alloc();\
197: (d)=_nd_free_list; _nd_free_list = (ND)BDY(_nd_free_list);\
198: NV(d)=(n); LEN(d)=(len); BDY(d)=(m)
1.40 noro 199: #define NEWNDV(d) ((d)=(NDV)MALLOC(sizeof(struct oNDV)))
200: #define MKNDV(n,m,l,d) NEWNDV(d); NV(d)=(n); BDY(d)=(m); LEN(d) = l;
1.63 noro 201: #define NEWNM_ind_pair(p)\
202: ((p)=(NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair)))
1.1 noro 203:
1.40 noro 204: /* allocate and link a new object */
1.13 noro 205: #define NEXTRHist(r,c) \
206: if(!(r)){NEWRHist(r);(c)=(r);}else{NEWRHist(NEXT(c));(c)=NEXT(c);}
1.1 noro 207: #define NEXTNM(r,c) \
208: if(!(r)){NEWNM(r);(c)=(r);}else{NEWNM(NEXT(c));(c)=NEXT(c);}
209: #define NEXTNM2(r,c,s) \
210: if(!(r)){(c)=(r)=(s);}else{NEXT(c)=(s);(c)=(s);}
1.40 noro 211: #define NEXTND_pairs(r,c) \
212: if(!(r)){NEWND_pairs(r);(c)=(r);}else{NEWND_pairs(NEXT(c));(c)=NEXT(c);}
1.63 noro 213: #define MKNM_ind_pair(p,m,i) (NEWNM_ind_pair(p),(p)->mul=(m),(p)->index=(i))
1.34 noro 214:
1.40 noro 215: /* deallocators */
1.1 noro 216: #define FREENM(m) NEXT(m)=_nm_free_list; _nm_free_list=(m)
217: #define FREENDP(m) NEXT(m)=_ndp_free_list; _ndp_free_list=(m)
218: #define FREEND(m) BDY(m)=(NM)_nd_free_list; _nd_free_list=(m)
219:
1.40 noro 220: /* macro for increasing pointer to NMV */
221: #define NMV_ADV(m) (m = (NMV)(((char *)m)+nmv_adv))
1.61 noro 222: #define NMV_OADV(m) (m = (NMV)(((char *)m)+oadv))
223: #define NDV_NADV(m) (m = (NMV)(((char *)m)+newadv))
1.56 noro 224: #define NMV_PREV(m) (m = (NMV)(((char *)m)-nmv_adv))
1.61 noro 225: #define NMV_OPREV(m) (m = (NMV)(((char *)m)-oadv))
226:
1.40 noro 227: /* external functions */
228: void GC_gcollect();
229: NODE append_one(NODE,int);
1.1 noro 230:
1.40 noro 231: /* manipulation of coefficients */
1.20 noro 232: void nd_removecont(int mod,ND p);
1.21 noro 233: void nd_removecont2(ND p1,ND p2);
1.40 noro 234: void removecont_array(Q *c,int n);
235:
236: /* GeoBucket functions */
1.25 noro 237: ND normalize_pbucket(int mod,PGeoBucket g);
238: int head_pbucket(int mod,PGeoBucket g);
1.26 noro 239: int head_pbucket_q(PGeoBucket g);
1.63 noro 240: void add_pbucket_symbolic(PGeoBucket g,ND d);
1.31 noro 241: void add_pbucket(int mod,PGeoBucket g,ND d);
1.25 noro 242: void free_pbucket(PGeoBucket b);
1.26 noro 243: void mulq_pbucket(PGeoBucket g,Q c);
1.63 noro 244: NM remove_head_pbucket_symbolic(PGeoBucket g);
1.25 noro 245: PGeoBucket create_pbucket();
1.20 noro 246:
1.40 noro 247: /* manipulation of pairs and bases */
1.39 noro 248: int nd_newps(int mod,ND a,ND aq);
1.40 noro 249: ND_pairs nd_newpairs( NODE g, int t );
1.1 noro 250: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest );
1.63 noro 251: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest );
1.1 noro 252: NODE update_base(NODE nd,int ndp);
1.40 noro 253: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t);
254: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest );
255: ND_pairs crit_B( ND_pairs d, int s );
256: ND_pairs crit_M( ND_pairs d1 );
257: ND_pairs crit_F( ND_pairs d1 );
1.1 noro 258: int crit_2( int dp1, int dp2 );
1.63 noro 259: int ndv_newps(NDV a,NDV aq);
1.40 noro 260:
261: /* top level functions */
1.63 noro 262: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp);
1.52 noro 263: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp);
1.63 noro 264: NODE nd_f4(int m);
1.27 noro 265: NODE nd_gb(int m,int checkonly);
1.23 noro 266: NODE nd_gb_trace(int m);
1.40 noro 267:
268: /* ndl functions */
1.61 noro 269: int ndl_weight(UINT *d);
270: void ndl_weight_mask(UINT *d);
271: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight);
272: void ndl_dehomogenize(UINT *p);
273: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos);
274: INLINE int ndl_reducible(UINT *d1,UINT *d2);
275: INLINE int ndl_lex_compare(UINT *d1,UINT *d2);
276: INLINE int ndl_block_compare(UINT *d1,UINT *d2);
277: INLINE int ndl_equal(UINT *d1,UINT *d2);
278: INLINE void ndl_copy(UINT *d1,UINT *d2);
279: INLINE void ndl_zero(UINT *d);
280: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d);
281: INLINE void ndl_addto(UINT *d1,UINT *d2);
282: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d);
283: INLINE int ndl_hash_value(UINT *d);
1.45 noro 284:
285: /* normal forms */
1.63 noro 286: INLINE int ndl_find_reducer(UINT *g);
287: INLINE int ndl_find_reducer_direct(UINT *g,NDV *ps,int len);
1.53 noro 288: int nd_sp(int mod,int trace,ND_pairs p,ND *nf);
289: int nd_nf(int mod,ND g,NDV *ps,int full,ND *nf);
290: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *nf);
1.45 noro 291: int nd_nf_direct(int mod,ND g,BaseSet base,int full,ND *rp);
1.40 noro 292:
293: /* finalizers */
1.61 noro 294: NODE ndv_reducebase(NODE x);
295: NODE ndv_reduceall(int m,NODE f);
1.40 noro 296:
297: /* allocators */
298: void nd_free_private_storage();
299: void _NM_alloc();
300: void _ND_alloc();
1.1 noro 301: void nd_free(ND p);
1.40 noro 302: void nd_free_redlist();
303:
304: /* printing */
1.61 noro 305: void ndl_print(UINT *dl);
1.1 noro 306: void nd_print(ND p);
1.16 noro 307: void nd_print_q(ND p);
1.1 noro 308: void ndp_print(ND_pairs d);
1.40 noro 309:
310:
311: /* setup, reconstruct */
312: void nd_init_ord(struct order_spec *spec);
313: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs ndp);
314: void nd_reconstruct_direct(int mod,NDV *ps,int len);
1.61 noro 315: void ndv_setup(int mod,int trace,NODE f);
316: void nd_setup_parameters(int nvar,int max);
1.43 noro 317: BlockMask nd_create_blockmask(struct order_spec *ord);
1.57 noro 318: EPOS nd_create_epos(struct order_spec *ord);
1.48 noro 319: int nd_get_exporigin(struct order_spec *ord);
1.61 noro 320: void ndv_mod(int mod,NDV p);
321: NDV ndv_dup(int mod,NDV p);
1.63 noro 322: ND nd_dup(ND p);
1.40 noro 323:
324: /* ND functions */
1.61 noro 325: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand);
1.40 noro 326: void nd_mul_c(int mod,ND p,int mul);
327: void nd_mul_c_q(ND p,Q mul);
1.61 noro 328: void nd_mul_c_p(VL vl,ND p,P mul);
1.40 noro 329: ND nd_remove_head(ND p);
1.1 noro 330: int nd_length(ND p);
1.61 noro 331: void nd_append_red(UINT *d,int i);
332: UINT *ndv_compute_bound(NDV p);
1.6 noro 333: ND nd_copy(ND p);
1.63 noro 334: ND nd_merge(ND p1,ND p2);
1.40 noro 335: ND nd_add(int mod,ND p1,ND p2);
336: ND nd_add_q(ND p1,ND p2);
1.41 noro 337: INLINE int nd_length(ND p);
1.4 noro 338:
1.40 noro 339: /* NDV functions */
1.55 noro 340: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p);
341: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen);
1.19 noro 342: void ndv_mul_c(int mod,NDV p,int mul);
1.40 noro 343: void ndv_mul_c_q(NDV p,Q mul);
1.63 noro 344: ND ndv_mul_nm_symbolic(NM m0,NDV p);
1.61 noro 345: ND ndv_mul_nm(int mod,NM m0,NDV p);
1.43 noro 346: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos);
1.61 noro 347: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos);
348: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS eops);
1.45 noro 349: void ndv_dehomogenize(NDV p,struct order_spec *spec);
1.40 noro 350: void ndv_removecont(int mod,NDV p);
351: void ndv_print(NDV p);
352: void ndv_print_q(NDV p);
353: void ndv_free(NDV p);
354:
355: /* converters */
1.61 noro 356: ND ptond(VL vl,VL dvl,P p);
357: NDV ptondv(VL vl,VL dvl,P p);
358: P ndvtop(int mod,VL vl,VL dvl,NDV p);
1.16 noro 359: NDV ndtondv(int mod,ND p);
1.23 noro 360: ND ndvtond(int mod,NDV p);
1.63 noro 361: int nm_ind_pair_to_vect(int m,UINT *s0,int n,NM_ind_pair pair,UINT *r);
1.67 noro 362: IndArray nm_ind_pair_to_vect_compress(int m,UINT *s0,int n,NM_ind_pair pair);
1.63 noro 363: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r);
1.1 noro 364:
365: void nd_free_private_storage()
366: {
367: _nd_free_list = 0;
368: _nm_free_list = 0;
1.5 noro 369: _ndp_free_list = 0;
1.13 noro 370: bzero(nd_red,sizeof(REDTAB_LEN*sizeof(RHist)));
1.1 noro 371: GC_gcollect();
372: }
373:
374: void _NM_alloc()
375: {
376: NM p;
377: int i;
378:
1.11 noro 379: for ( i = 0; i < 1024; i++ ) {
1.61 noro 380: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 381: p->next = _nm_free_list; _nm_free_list = p;
382: }
383: }
384:
385: void _ND_alloc()
386: {
387: ND p;
388: int i;
389:
390: for ( i = 0; i < 1024; i++ ) {
391: p = (ND)GC_malloc(sizeof(struct oND));
392: p->body = (NM)_nd_free_list; _nd_free_list = p;
393: }
394: }
395:
396: void _NDP_alloc()
397: {
398: ND_pairs p;
399: int i;
400:
1.11 noro 401: for ( i = 0; i < 1024; i++ ) {
1.1 noro 402: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 403: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 404: p->next = _ndp_free_list; _ndp_free_list = p;
405: }
406: }
407:
1.30 noro 408: INLINE int nd_length(ND p)
1.1 noro 409: {
410: NM m;
411: int i;
412:
413: if ( !p )
414: return 0;
415: else {
416: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
417: return i;
418: }
419: }
420:
1.61 noro 421: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 422: {
1.61 noro 423: UINT u1,u2;
1.1 noro 424: int i,j;
425:
1.34 noro 426: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 427: #if USE_UNROLL
1.1 noro 428: switch ( nd_bpe ) {
1.62 noro 429: case 3:
430: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
431: u1 = d1[i]; u2 = d2[i];
432: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
433: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
434: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
435: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
436: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
437: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
438: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
439: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
440: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
441: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
442: }
443: return 1;
444: break;
1.1 noro 445: case 4:
1.41 noro 446: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 447: u1 = d1[i]; u2 = d2[i];
448: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 449: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
450: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
451: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
452: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
453: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
454: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
455: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 456: }
457: return 1;
458: break;
459: case 6:
1.41 noro 460: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 461: u1 = d1[i]; u2 = d2[i];
462: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 463: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
464: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
465: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
466: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 467: }
468: return 1;
469: break;
470: case 8:
1.41 noro 471: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 472: u1 = d1[i]; u2 = d2[i];
473: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 474: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
475: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
476: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 477: }
478: return 1;
479: break;
480: case 16:
1.41 noro 481: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 482: u1 = d1[i]; u2 = d2[i];
483: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 484: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 485: }
486: return 1;
487: break;
488: case 32:
1.41 noro 489: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 490: if ( d1[i] < d2[i] ) return 0;
491: return 1;
492: break;
493: default:
1.41 noro 494: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 495: u1 = d1[i]; u2 = d2[i];
496: for ( j = 0; j < nd_epw; j++ )
497: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
498: }
499: return 1;
500: }
1.65 noro 501: #else
502: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
503: u1 = d1[i]; u2 = d2[i];
504: for ( j = 0; j < nd_epw; j++ )
505: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
506: }
507: return 1;
508: #endif
1.1 noro 509: }
510:
1.61 noro 511: /*
512: * If the current order is a block order,
513: * then the last block is length 1 and contains
514: * the homo variable. Otherwise, the original
515: * order is either 0 or 2.
516: */
517:
518: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 519: {
1.61 noro 520: int w,i,e,n,omask0;
521:
522: omask0 = (1<<obpe)-1;
523: n = nd_nvar-1;
524: ndl_zero(r);
525: for ( i = 0; i < n; i++ ) {
526: e = GET_EXP_OLD(d,i);
527: PUT_EXP(r,i,e);
528: }
529: w = TD(d);
530: PUT_EXP(r,nd_nvar-1,weight-w);
531: TD(r) = weight;
532: if ( nd_blockmask ) ndl_weight_mask(r);
533: }
534:
535: void ndl_dehomogenize(UINT *d)
536: {
537: UINT mask;
538: UINT h;
1.31 noro 539: int i,bits;
1.23 noro 540:
1.44 noro 541: if ( nd_blockmask ) {
542: h = GET_EXP(d,nd_nvar-1);
1.45 noro 543: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 544: TD(d) -= h;
545: d[nd_exporigin-1] -= h;
546: } else {
547: if ( nd_isrlex ) {
548: if ( nd_bpe == 32 ) {
549: h = d[nd_exporigin];
550: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
551: d[i-1] = d[i];
552: d[i-1] = 0;
553: TD(d) -= h;
554: } else {
555: bits = nd_epw*nd_bpe;
556: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
557: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
558: for ( i = nd_exporigin; i < nd_wpd; i++ )
559: d[i] = ((d[i]<<nd_bpe)&mask)
560: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
561: TD(d) -= h;
562: }
1.45 noro 563: } else {
564: h = GET_EXP(d,nd_nvar-1);
565: XOR_EXP(d,nd_nvar-1,h);
566: TD(d) -= h;
567: }
1.44 noro 568: }
1.23 noro 569: }
570:
1.61 noro 571: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 572: {
1.61 noro 573: UINT t1,t2,u,u1,u2;
1.43 noro 574: int i,j,l;
1.1 noro 575:
1.65 noro 576: #if USE_UNROLL
1.1 noro 577: switch ( nd_bpe ) {
1.62 noro 578: case 3:
579: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
580: u1 = d1[i]; u2 = d2[i];
581: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
582: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
583: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
584: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
585: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
586: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
587: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
588: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
589: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
590: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
591: d[i] = u;
592: }
593: break;
1.1 noro 594: case 4:
1.41 noro 595: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 596: u1 = d1[i]; u2 = d2[i];
597: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 598: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
599: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
600: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
601: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
602: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
603: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
604: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 605: d[i] = u;
606: }
607: break;
608: case 6:
1.41 noro 609: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 610: u1 = d1[i]; u2 = d2[i];
611: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 612: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
613: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
614: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
615: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 616: d[i] = u;
617: }
618: break;
619: case 8:
1.41 noro 620: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 621: u1 = d1[i]; u2 = d2[i];
622: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 623: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
624: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
625: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 626: d[i] = u;
627: }
628: break;
629: case 16:
1.41 noro 630: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 631: u1 = d1[i]; u2 = d2[i];
632: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 633: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 634: d[i] = u;
635: }
636: break;
637: case 32:
1.41 noro 638: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 639: u1 = d1[i]; u2 = d2[i];
640: d[i] = u1>u2?u1:u2;
641: }
642: break;
643: default:
1.41 noro 644: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 645: u1 = d1[i]; u2 = d2[i];
646: for ( j = 0, u = 0; j < nd_epw; j++ ) {
647: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
648: }
649: d[i] = u;
650: }
651: break;
652: }
1.65 noro 653: #else
654: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
655: u1 = d1[i]; u2 = d2[i];
656: for ( j = 0, u = 0; j < nd_epw; j++ ) {
657: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
658: }
659: d[i] = u;
660: }
661: #endif
1.39 noro 662: TD(d) = ndl_weight(d);
1.61 noro 663: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 664: }
665:
1.61 noro 666: int ndl_weight(UINT *d)
1.1 noro 667: {
1.61 noro 668: UINT t,u;
1.1 noro 669: int i,j;
670:
1.60 noro 671: if ( current_dl_weight_vector )
672: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
673: u = GET_EXP(d,i);
674: t += MUL_WEIGHT(u,i);
675: }
676: else
677: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
678: u = d[i];
679: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
680: t += (u&nd_mask0);
681: }
1.1 noro 682: return t;
683: }
684:
1.61 noro 685: void ndl_weight_mask(UINT *d)
1.43 noro 686: {
1.61 noro 687: UINT t,u;
688: UINT *mask;
689: int i,j,k,l;
1.43 noro 690:
1.61 noro 691: l = nd_blockmask->n;
692: for ( k = 0; k < l; k++ ) {
693: mask = nd_blockmask->mask[k];
694: if ( current_dl_weight_vector )
695: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
696: u = GET_EXP_MASK(d,i,mask);
697: t += MUL_WEIGHT(u,i);
698: }
699: else
700: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
701: u = d[i]&mask[i];
702: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
703: t += (u&nd_mask0);
704: }
705: d[k+1] = t;
706: }
1.43 noro 707: }
708:
1.61 noro 709: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 710: {
711: int i;
712:
1.41 noro 713: d1 += nd_exporigin;
714: d2 += nd_exporigin;
715: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 716: if ( *d1 > *d2 )
1.32 noro 717: return nd_isrlex ? -1 : 1;
1.1 noro 718: else if ( *d1 < *d2 )
1.32 noro 719: return nd_isrlex ? 1 : -1;
1.1 noro 720: return 0;
721: }
722:
1.61 noro 723: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 724: {
725: int i,l,j,ord_o,ord_l;
726: struct order_pair *op;
1.61 noro 727: UINT t1,t2,m;
728: UINT *mask;
1.43 noro 729:
730: l = nd_blockmask->n;
731: op = nd_blockmask->order_pair;
732: for ( j = 0; j < l; j++ ) {
733: mask = nd_blockmask->mask[j];
734: ord_o = op[j].order;
735: if ( ord_o < 2 )
1.44 noro 736: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
737: else if ( t1 < t2 ) return -1;
1.43 noro 738: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 739: m = mask[i];
740: t1 = d1[i]&m;
741: t2 = d2[i]&m;
1.43 noro 742: if ( t1 > t2 )
743: return !ord_o ? -1 : 1;
744: else if ( t1 < t2 )
745: return !ord_o ? 1 : -1;
746: }
747: }
748: return 0;
749: }
750:
1.58 noro 751: /* TDH -> WW -> TD-> RL */
752:
1.61 noro 753: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 754: {
755: int i,m,e1,e2;
756:
757: if ( TD(d1) > TD(d2) ) return 1;
758: else if ( TD(d1) < TD(d2) ) return -1;
759: m = nd_nvar>>1;
760: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
761: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
762: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
763: }
764: if ( e1 > e2 ) return 1;
765: else if ( e1 < e2 ) return -1;
766: return ndl_lex_compare(d1,d2);
767: }
768:
1.61 noro 769: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 770: {
771: int i;
772:
1.41 noro 773: for ( i = 0; i < nd_wpd; i++ )
1.34 noro 774: if ( *d1++ != *d2++ )
1.1 noro 775: return 0;
776: return 1;
777: }
778:
1.61 noro 779: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 780: {
781: int i;
782:
783: switch ( nd_wpd ) {
1.41 noro 784: case 2:
1.34 noro 785: TD(d2) = TD(d1);
786: d2[1] = d1[1];
1.6 noro 787: break;
1.41 noro 788: case 3:
1.34 noro 789: TD(d2) = TD(d1);
1.6 noro 790: d2[1] = d1[1];
1.34 noro 791: d2[2] = d1[2];
1.6 noro 792: break;
793: default:
1.41 noro 794: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 795: d2[i] = d1[i];
796: break;
797: }
798: }
799:
1.61 noro 800: INLINE void ndl_zero(UINT *d)
801: {
802: int i;
803: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
804: }
805:
806: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 807: {
808: int i;
809:
1.43 noro 810: #if 1
1.6 noro 811: switch ( nd_wpd ) {
1.41 noro 812: case 2:
813: TD(d) = TD(d1)+TD(d2);
1.34 noro 814: d[1] = d1[1]+d2[1];
1.6 noro 815: break;
1.41 noro 816: case 3:
817: TD(d) = TD(d1)+TD(d2);
1.6 noro 818: d[1] = d1[1]+d2[1];
1.34 noro 819: d[2] = d1[2]+d2[2];
1.6 noro 820: break;
821: default:
1.43 noro 822: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 823: break;
824: }
1.43 noro 825: #else
826: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
827: #endif
1.6 noro 828: }
829:
1.55 noro 830: /* d1 += d2 */
1.61 noro 831: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 832: {
833: int i;
834:
835: #if 1
836: switch ( nd_wpd ) {
837: case 2:
838: TD(d1) += TD(d2);
839: d1[1] += d2[1];
840: break;
841: case 3:
842: TD(d1) += TD(d2);
843: d1[1] += d2[1];
844: d1[2] += d2[2];
845: break;
846: default:
847: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
848: break;
849: }
850: #else
851: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
852: #endif
853: }
854:
1.61 noro 855: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 856: {
857: int i;
858:
1.43 noro 859: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 860: }
861:
1.61 noro 862: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 863: {
1.61 noro 864: UINT t1,t2,u,u1,u2;
1.1 noro 865: int i,j;
866:
1.65 noro 867: #if USE_UNROLL
1.1 noro 868: switch ( nd_bpe ) {
1.62 noro 869: case 3:
870: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
871: u1 = d1[i]; u2 = d2[i];
872: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
873: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
874: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
875: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
876: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
877: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
878: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
879: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
880: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
881: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
882: }
883: return 1;
884: break;
1.1 noro 885: case 4:
1.41 noro 886: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 887: u1 = d1[i]; u2 = d2[i];
888: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 889: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
890: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
891: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
892: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
893: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
894: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
895: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 896: }
897: return 1;
898: break;
899: case 6:
1.41 noro 900: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 901: u1 = d1[i]; u2 = d2[i];
902: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 903: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
904: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
905: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
906: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 907: }
908: return 1;
909: break;
910: case 8:
1.41 noro 911: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 912: u1 = d1[i]; u2 = d2[i];
913: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 914: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
915: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
916: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 917: }
918: return 1;
919: break;
920: case 16:
1.41 noro 921: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 922: u1 = d1[i]; u2 = d2[i];
923: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 924: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 925: }
926: return 1;
927: break;
928: case 32:
1.41 noro 929: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 930: if ( d1[i] && d2[i] ) return 0;
931: return 1;
932: break;
933: default:
1.41 noro 934: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 935: u1 = d1[i]; u2 = d2[i];
936: for ( j = 0; j < nd_epw; j++ ) {
937: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
938: u1 >>= nd_bpe; u2 >>= nd_bpe;
939: }
940: }
941: return 1;
942: break;
943: }
1.65 noro 944: #else
945: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
946: u1 = d1[i]; u2 = d2[i];
947: for ( j = 0; j < nd_epw; j++ ) {
948: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
949: u1 >>= nd_bpe; u2 >>= nd_bpe;
950: }
951: }
952: return 1;
953: #endif
1.1 noro 954: }
955:
1.61 noro 956: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 957: {
1.61 noro 958: UINT u2;
959: UINT *d1;
1.5 noro 960: int i,j,ind,k;
1.1 noro 961:
1.5 noro 962: d1 = nd_bound[index];
963: ind = 0;
1.65 noro 964: #if USE_UNROLL
1.5 noro 965: switch ( nd_bpe ) {
1.62 noro 966: case 3:
967: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
968: u2 = d2[i];
969: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
970: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
971: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
972: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
973: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
974: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
975: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
976: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
977: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
978: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
979: }
980: return 0;
981: break;
1.5 noro 982: case 4:
1.41 noro 983: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 984: u2 = d2[i];
985: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
986: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
987: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
988: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
989: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
990: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
991: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
992: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
993: }
994: return 0;
995: break;
996: case 6:
1.41 noro 997: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 998: u2 = d2[i];
999: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1000: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1001: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1002: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1003: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1004: }
1005: return 0;
1006: break;
1007: case 8:
1.41 noro 1008: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1009: u2 = d2[i];
1010: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1011: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1012: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1013: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1014: }
1015: return 0;
1016: break;
1017: case 16:
1.41 noro 1018: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1019: u2 = d2[i];
1020: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1021: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1022: }
1023: return 0;
1024: break;
1025: case 32:
1.41 noro 1026: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1027: if ( d1[i]+d2[i]<d1[i] ) return 1;
1028: return 0;
1029: break;
1030: default:
1.41 noro 1031: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1032: u2 = d2[i];
1033: k = (nd_epw-1)*nd_bpe;
1034: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1035: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1036: }
1037: return 0;
1038: break;
1039: }
1.65 noro 1040: #else
1041: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1042: u2 = d2[i];
1043: k = (nd_epw-1)*nd_bpe;
1044: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1045: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1046: }
1047: return 0;
1048: #endif
1.1 noro 1049: }
1050:
1.61 noro 1051: int ndl_check_bound2_direct(UINT *d1,UINT *d2)
1.23 noro 1052: {
1.61 noro 1053: UINT u2;
1.45 noro 1054: int i,j,ind,k;
1.23 noro 1055:
1.45 noro 1056: ind = 0;
1.65 noro 1057: #if USE_UNROLL
1.23 noro 1058: switch ( nd_bpe ) {
1.62 noro 1059: case 3:
1060: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1061: u2 = d2[i];
1062: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
1063: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
1064: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
1065: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
1066: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
1067: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
1068: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
1069: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
1070: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
1071: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
1072: }
1073: return 0;
1074: break;
1.23 noro 1075: case 4:
1.41 noro 1076: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.45 noro 1077: u2 = d2[i];
1078: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
1079: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
1080: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
1081: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
1082: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
1083: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
1084: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1085: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1.23 noro 1086: }
1087: return 0;
1088: break;
1089: case 6:
1.41 noro 1090: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.45 noro 1091: u2 = d2[i];
1092: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1093: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1094: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1095: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1096: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1.23 noro 1097: }
1098: return 0;
1099: break;
1100: case 8:
1.41 noro 1101: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.45 noro 1102: u2 = d2[i];
1103: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1104: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1105: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1106: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1.23 noro 1107: }
1108: return 0;
1109: break;
1110: case 16:
1.41 noro 1111: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.45 noro 1112: u2 = d2[i];
1113: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1114: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1.23 noro 1115: }
1116: return 0;
1117: break;
1118: case 32:
1.41 noro 1119: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.23 noro 1120: if ( d1[i]+d2[i]<d1[i] ) return 1;
1121: return 0;
1122: break;
1123: default:
1.41 noro 1124: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.45 noro 1125: u2 = d2[i];
1.23 noro 1126: k = (nd_epw-1)*nd_bpe;
1127: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1.45 noro 1128: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1.23 noro 1129: }
1130: return 0;
1131: break;
1132: }
1.65 noro 1133: #else
1134: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1135: u2 = d2[i];
1136: k = (nd_epw-1)*nd_bpe;
1137: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1138: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1139: }
1140: return 0;
1141: #endif
1.23 noro 1142: }
1143:
1.61 noro 1144: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1145: {
1146: int i;
1147: int r;
1148:
1.34 noro 1149: r = 0;
1.41 noro 1150: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1151: r = ((r<<16)+d[i])%REDTAB_LEN;
1152: return r;
1153: }
1154:
1.63 noro 1155: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1156: {
1.13 noro 1157: RHist r;
1.6 noro 1158: int d,k,i;
1.1 noro 1159:
1.63 noro 1160: d = ndl_hash_value(dg);
1.13 noro 1161: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1162: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1163: if ( k > 0 ) nd_notfirst++;
1164: nd_found++;
1.13 noro 1165: return r->index;
1.1 noro 1166: }
1167: }
1.13 noro 1168: if ( Reverse )
1169: for ( i = nd_psn-1; i >= 0; i-- ) {
1170: r = nd_psh[i];
1.34 noro 1171: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1172: nd_create++;
1.34 noro 1173: nd_append_red(dg,i);
1.13 noro 1174: return i;
1175: }
1176: }
1177: else
1178: for ( i = 0; i < nd_psn; i++ ) {
1179: r = nd_psh[i];
1.34 noro 1180: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1181: nd_create++;
1.34 noro 1182: nd_append_red(dg,i);
1.13 noro 1183: return i;
1184: }
1.1 noro 1185: }
1.6 noro 1186: return -1;
1.1 noro 1187: }
1188:
1.63 noro 1189: INLINE int ndl_find_reducer_direct(UINT *dg,NDV *ps,int len)
1.23 noro 1190: {
1191: NDV r;
1.31 noro 1192: RHist s;
1193: int d,k,i;
1.23 noro 1194:
1195: if ( Reverse )
1196: for ( i = len-1; i >= 0; i-- ) {
1197: r = ps[i];
1.63 noro 1198: if ( ndl_reducible(dg,HDL(r)) )
1.23 noro 1199: return i;
1200: }
1201: else
1202: for ( i = 0; i < len; i++ ) {
1203: r = ps[i];
1.63 noro 1204: if ( ndl_reducible(dg,HDL(r)) )
1.23 noro 1205: return i;
1206: }
1207: return -1;
1208: }
1209:
1.63 noro 1210: ND nd_merge(ND p1,ND p2)
1211: {
1212: int n,c;
1213: int t,can,td1,td2;
1214: ND r;
1215: NM m1,m2,mr0,mr,s;
1216:
1217: if ( !p1 ) return p2;
1218: else if ( !p2 ) return p1;
1219: else {
1220: can = 0;
1221: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1222: c = DL_COMPARE(DL(m1),DL(m2));
1223: switch ( c ) {
1224: case 0:
1225: s = m1; m1 = NEXT(m1);
1226: can++; NEXTNM2(mr0,mr,s);
1227: s = m2; m2 = NEXT(m2); FREENM(s);
1228: break;
1229: case 1:
1230: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1231: break;
1232: case -1:
1233: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1234: break;
1235: }
1236: }
1237: if ( !mr0 )
1238: if ( m1 ) mr0 = m1;
1239: else if ( m2 ) mr0 = m2;
1240: else return 0;
1241: else if ( m1 ) NEXT(mr) = m1;
1242: else if ( m2 ) NEXT(mr) = m2;
1243: else NEXT(mr) = 0;
1244: BDY(p1) = mr0;
1245: SG(p1) = MAX(SG(p1),SG(p2));
1246: LEN(p1) = LEN(p1)+LEN(p2)-can;
1247: FREEND(p2);
1248: return p1;
1249: }
1250: }
1251:
1.31 noro 1252: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1253: {
1254: int n,c;
1.34 noro 1255: int t,can,td1,td2;
1.1 noro 1256: ND r;
1257: NM m1,m2,mr0,mr,s;
1258:
1.34 noro 1259: if ( !p1 ) return p2;
1260: else if ( !p2 ) return p1;
1261: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1262: else {
1.30 noro 1263: can = 0;
1.1 noro 1264: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1265: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1266: switch ( c ) {
1267: case 0:
1.19 noro 1268: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1269: if ( t < 0 ) t += mod;
1.1 noro 1270: s = m1; m1 = NEXT(m1);
1271: if ( t ) {
1.34 noro 1272: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1273: } else {
1.34 noro 1274: can += 2; FREENM(s);
1.1 noro 1275: }
1276: s = m2; m2 = NEXT(m2); FREENM(s);
1277: break;
1278: case 1:
1279: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1280: break;
1281: case -1:
1282: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1283: break;
1284: }
1285: }
1286: if ( !mr0 )
1.34 noro 1287: if ( m1 ) mr0 = m1;
1288: else if ( m2 ) mr0 = m2;
1289: else return 0;
1290: else if ( m1 ) NEXT(mr) = m1;
1291: else if ( m2 ) NEXT(mr) = m2;
1292: else NEXT(mr) = 0;
1.1 noro 1293: BDY(p1) = mr0;
1.14 noro 1294: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1295: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1296: FREEND(p2);
1297: return p1;
1298: }
1299: }
1300:
1.31 noro 1301: ND nd_add_q(ND p1,ND p2)
1.17 noro 1302: {
1.30 noro 1303: int n,c,can;
1.17 noro 1304: ND r;
1305: NM m1,m2,mr0,mr,s;
1306: Q t;
1307:
1.34 noro 1308: if ( !p1 ) return p2;
1309: else if ( !p2 ) return p1;
1.31 noro 1310: else {
1.30 noro 1311: can = 0;
1.17 noro 1312: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1313: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1314: switch ( c ) {
1315: case 0:
1316: addq(CQ(m1),CQ(m2),&t);
1317: s = m1; m1 = NEXT(m1);
1318: if ( t ) {
1.34 noro 1319: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1320: } else {
1.34 noro 1321: can += 2; FREENM(s);
1.17 noro 1322: }
1323: s = m2; m2 = NEXT(m2); FREENM(s);
1324: break;
1325: case 1:
1326: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1327: break;
1328: case -1:
1329: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1330: break;
1331: }
1332: }
1333: if ( !mr0 )
1.34 noro 1334: if ( m1 ) mr0 = m1;
1335: else if ( m2 ) mr0 = m2;
1336: else return 0;
1337: else if ( m1 ) NEXT(mr) = m1;
1338: else if ( m2 ) NEXT(mr) = m2;
1339: else NEXT(mr) = 0;
1.17 noro 1340: BDY(p1) = mr0;
1341: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1342: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1343: FREEND(p2);
1344: return p1;
1345: }
1346: }
1347:
1.1 noro 1348: /* ret=1 : success, ret=0 : overflow */
1.53 noro 1349: int nd_nf(int mod,ND g,NDV *ps,int full,ND *rp)
1.1 noro 1350: {
1.11 noro 1351: ND d;
1.1 noro 1352: NM m,mrd,tail;
1.7 noro 1353: NM mul;
1.10 noro 1354: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1355: int c,c1,c2,dummy;
1.17 noro 1356: RHist h;
1.11 noro 1357: NDV p,red;
1.16 noro 1358: Q cg,cred,gcd;
1.21 noro 1359: double hmag;
1.1 noro 1360:
1361: if ( !g ) {
1362: *rp = 0;
1363: return 1;
1364: }
1.34 noro 1365: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1366:
1.14 noro 1367: sugar0 = sugar = SG(g);
1.1 noro 1368: n = NV(g);
1.61 noro 1369: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1370: for ( d = 0; g; ) {
1.63 noro 1371: index = ndl_find_reducer(HDL(g));
1.6 noro 1372: if ( index >= 0 ) {
1.17 noro 1373: h = nd_psh[index];
1374: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1375: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1376: nd_free(g); nd_free(d);
1377: return 0;
1378: }
1.53 noro 1379: p = ps[index];
1.16 noro 1380: if ( mod ) {
1.19 noro 1381: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1382: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1383: } else {
1.17 noro 1384: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1385: chsgnq(cg,&CQ(mul));
1.20 noro 1386: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.16 noro 1387: }
1.55 noro 1388: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1389: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1390: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.21 noro 1391: nd_removecont2(d,g);
1392: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1393: }
1.1 noro 1394: } else if ( !full ) {
1395: *rp = g;
1396: return 1;
1397: } else {
1398: m = BDY(g);
1399: if ( NEXT(m) ) {
1.34 noro 1400: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1401: } else {
1402: FREEND(g); g = 0;
1403: }
1404: if ( d ) {
1.34 noro 1405: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1406: } else {
1.34 noro 1407: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1408: }
1409: }
1410: }
1.34 noro 1411: if ( d ) SG(d) = sugar;
1.1 noro 1412: *rp = d;
1413: return 1;
1414: }
1.28 noro 1415:
1.53 noro 1416: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1417: {
1418: int hindex,index;
1419: NDV p;
1420: ND u,d,red;
1421: NODE l;
1.31 noro 1422: NM mul,m,mrd,tail;
1.25 noro 1423: int sugar,psugar,n,h_reducible;
1424: PGeoBucket bucket;
1425: int c,c1,c2;
1.26 noro 1426: Q cg,cred,gcd,zzz;
1.25 noro 1427: RHist h;
1.28 noro 1428: double hmag,gmag;
1.25 noro 1429:
1430: if ( !g ) {
1431: *rp = 0;
1432: return 1;
1433: }
1434: sugar = SG(g);
1435: n = NV(g);
1.34 noro 1436: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1437: bucket = create_pbucket();
1.31 noro 1438: add_pbucket(mod,bucket,g);
1.25 noro 1439: d = 0;
1.61 noro 1440: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1441: while ( 1 ) {
1.26 noro 1442: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1443: if ( hindex < 0 ) {
1.34 noro 1444: if ( d ) SG(d) = sugar;
1.25 noro 1445: *rp = d;
1446: return 1;
1447: }
1448: g = bucket->body[hindex];
1.63 noro 1449: index = ndl_find_reducer(HDL(g));
1.25 noro 1450: if ( index >= 0 ) {
1451: h = nd_psh[index];
1452: ndl_sub(HDL(g),DL(h),DL(mul));
1453: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1454: nd_free(d);
1.25 noro 1455: free_pbucket(bucket);
1456: *rp = 0;
1457: return 0;
1458: }
1.53 noro 1459: p = ps[index];
1.25 noro 1460: if ( mod ) {
1461: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1462: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1463: } else {
1464: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1465: chsgnq(cg,&CQ(mul));
1.26 noro 1466: nd_mul_c_q(d,cred);
1467: mulq_pbucket(bucket,cred);
1468: g = bucket->body[hindex];
1.28 noro 1469: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1470: }
1.55 noro 1471: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1472: bucket->body[hindex] = nd_remove_head(g);
1473: red = nd_remove_head(red);
1.31 noro 1474: add_pbucket(mod,bucket,red);
1.34 noro 1475: psugar = SG(p)+TD(DL(mul));
1476: sugar = MAX(sugar,psugar);
1.28 noro 1477: if ( !mod && hmag && (gmag > hmag) ) {
1478: g = normalize_pbucket(mod,bucket);
1479: if ( !g ) {
1.34 noro 1480: if ( d ) SG(d) = sugar;
1.28 noro 1481: *rp = d;
1482: return 1;
1483: }
1484: nd_removecont2(d,g);
1485: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1486: add_pbucket(mod,bucket,g);
1.28 noro 1487: }
1.25 noro 1488: } else if ( !full ) {
1489: g = normalize_pbucket(mod,bucket);
1.34 noro 1490: if ( g ) SG(g) = sugar;
1.25 noro 1491: *rp = g;
1492: return 1;
1493: } else {
1494: m = BDY(g);
1495: if ( NEXT(m) ) {
1.34 noro 1496: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1497: } else {
1498: FREEND(g); g = 0;
1499: }
1500: bucket->body[hindex] = g;
1501: NEXT(m) = 0;
1502: if ( d ) {
1.34 noro 1503: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1504: } else {
1.34 noro 1505: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1506: }
1507: }
1508: }
1509: }
1.27 noro 1510:
1.45 noro 1511: int nd_nf_direct(int mod,ND g,BaseSet base,int full,ND *rp)
1.23 noro 1512: {
1.30 noro 1513: ND d;
1514: NM m,mrd,tail;
1515: NM mul;
1.45 noro 1516: NDV *ps;
1517: int n,sugar,psugar,sugar0,stat,index,len;
1.31 noro 1518: int c,c1,c2;
1.61 noro 1519: UINT **bound;
1.30 noro 1520: RHist h;
1521: NDV p,red;
1522: Q cg,cred,gcd;
1523: double hmag;
1524:
1525: if ( !g ) {
1526: *rp = 0;
1527: return 1;
1528: }
1529: #if 0
1530: if ( !mod )
1531: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1532: #else
1533: /* XXX */
1534: hmag = 0;
1535: #endif
1536:
1.45 noro 1537: ps = base->ps;
1538: bound = base->bound;
1539: len = base->len;
1.30 noro 1540: sugar0 = sugar = SG(g);
1541: n = NV(g);
1.61 noro 1542: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.30 noro 1543: for ( d = 0; g; ) {
1.63 noro 1544: index = ndl_find_reducer_direct(HDL(g),ps,len);
1.30 noro 1545: if ( index >= 0 ) {
1546: p = ps[index];
1547: ndl_sub(HDL(g),HDL(p),DL(mul));
1.45 noro 1548: if ( ndl_check_bound2_direct(bound[index],DL(mul)) ) {
1.30 noro 1549: nd_free(g); nd_free(d);
1550: return 0;
1551: }
1552: if ( mod ) {
1553: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1554: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1555: } else {
1556: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1557: chsgnq(cg,&CQ(mul));
1558: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1559: }
1.55 noro 1560: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1561: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.30 noro 1562: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1563: nd_removecont2(d,g);
1564: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1565: }
1566: } else if ( !full ) {
1567: *rp = g;
1568: return 1;
1569: } else {
1570: m = BDY(g);
1571: if ( NEXT(m) ) {
1.34 noro 1572: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.30 noro 1573: } else {
1574: FREEND(g); g = 0;
1575: }
1576: if ( d ) {
1.34 noro 1577: NEXT(tail)=m; tail=m; LEN(d)++;
1.30 noro 1578: } else {
1.34 noro 1579: MKND(n,m,1,d); tail = BDY(d);
1.30 noro 1580: }
1581: }
1582: }
1.34 noro 1583: if ( d ) SG(d) = sugar;
1.30 noro 1584: *rp = d;
1585: return 1;
1586: }
1587:
1.61 noro 1588: /* input : list of NDV, cand : list of NDV */
1.28 noro 1589:
1.61 noro 1590: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1591: {
1592: int n,i,stat;
1593: ND nf,d;
1.61 noro 1594: NDV r;
1.45 noro 1595: NODE t,s;
1596:
1.61 noro 1597: ndv_setup(0,0,cand);
1.31 noro 1598: n = length(cand);
1.28 noro 1599:
1600: /* membercheck : list is a subset of Id(cand) ? */
1601: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1602: again:
1.61 noro 1603: if ( nd_bpe > obpe )
1604: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1605: else
1606: r = (NDV)BDY(t);
1607: d = ndvtond(0,r);
1.53 noro 1608: stat = nd_nf(0,d,nd_ps,0,&nf);
1.45 noro 1609: if ( !stat ) {
1610: nd_reconstruct(0,0,0);
1611: goto again;
1612: } else if ( nf ) return 0;
1613: printf("."); fflush(stdout);
1.28 noro 1614: }
1.45 noro 1615: printf("\n");
1.28 noro 1616: /* gbcheck : cand is a GB of Id(cand) ? */
1.34 noro 1617: if ( !nd_gb(0,1) ) return 0;
1.28 noro 1618: /* XXX */
1.23 noro 1619: return 1;
1620: }
1.1 noro 1621:
1622: ND nd_remove_head(ND p)
1623: {
1624: NM m;
1625:
1626: m = BDY(p);
1627: if ( !NEXT(m) ) {
1.34 noro 1628: FREEND(p); p = 0;
1.31 noro 1629: } else {
1.34 noro 1630: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1631: }
1.1 noro 1632: FREENM(m);
1633: return p;
1634: }
1635:
1636: PGeoBucket create_pbucket()
1637: {
1638: PGeoBucket g;
1639:
1640: g = CALLOC(1,sizeof(struct oPGeoBucket));
1641: g->m = -1;
1642: return g;
1643: }
1644:
1.25 noro 1645: void free_pbucket(PGeoBucket b) {
1646: int i;
1647:
1.26 noro 1648: for ( i = 0; i <= b->m; i++ )
1.25 noro 1649: if ( b->body[i] ) {
1650: nd_free(b->body[i]);
1651: b->body[i] = 0;
1652: }
1653: GC_free(b);
1654: }
1655:
1.63 noro 1656: void add_pbucket_symbolic(PGeoBucket g,ND d)
1657: {
1658: int l,i,k,m;
1659:
1660: if ( !d )
1661: return;
1662: l = LEN(d);
1663: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1664: /* 2^(k-1) < l <= 2^k (=m) */
1665: d = nd_merge(g->body[k],d);
1666: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1667: g->body[k] = 0;
1668: d = nd_merge(g->body[k+1],d);
1669: }
1670: g->body[k] = d;
1671: g->m = MAX(g->m,k);
1672: }
1673:
1.31 noro 1674: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1675: {
1.31 noro 1676: int l,i,k,m;
1.1 noro 1677:
1.31 noro 1678: if ( !d )
1679: return;
1680: l = LEN(d);
1.29 noro 1681: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1682: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1683: d = nd_add(mod,g->body[k],d);
1684: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1685: g->body[k] = 0;
1.31 noro 1686: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1687: }
1688: g->body[k] = d;
1689: g->m = MAX(g->m,k);
1690: }
1691:
1.26 noro 1692: void mulq_pbucket(PGeoBucket g,Q c)
1693: {
1694: int k;
1695:
1696: for ( k = 0; k <= g->m; k++ )
1697: nd_mul_c_q(g->body[k],c);
1698: }
1699:
1.63 noro 1700: NM remove_head_pbucket_symbolic(PGeoBucket g)
1701: {
1702: int j,i,k,c;
1703: NM head;
1704:
1705: k = g->m;
1706: j = -1;
1707: for ( i = 0; i <= k; i++ ) {
1708: if ( !g->body[i] ) continue;
1709: if ( j < 0 ) j = i;
1710: else {
1711: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1712: if ( c > 0 )
1713: j = i;
1714: else if ( c == 0 )
1715: g->body[i] = nd_remove_head(g->body[i]);
1716: }
1717: }
1718: if ( j < 0 ) return 0;
1719: else {
1720: head = BDY(g->body[j]);
1721: if ( !NEXT(head) ) {
1722: FREEND(g->body[j]);
1723: g->body[j] = 0;
1724: } else {
1725: BDY(g->body[j]) = NEXT(head);
1726: LEN(g->body[j])--;
1727: }
1728: return head;
1729: }
1730: }
1731:
1.19 noro 1732: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1733: {
1734: int j,i,c,k,nv,sum;
1.61 noro 1735: UINT *di,*dj;
1.1 noro 1736: ND gi,gj;
1737:
1738: k = g->m;
1739: while ( 1 ) {
1740: j = -1;
1741: for ( i = 0; i <= k; i++ ) {
1742: if ( !(gi = g->body[i]) )
1743: continue;
1744: if ( j < 0 ) {
1745: j = i;
1746: gj = g->body[j];
1747: dj = HDL(gj);
1.14 noro 1748: sum = HCM(gj);
1.1 noro 1749: } else {
1.34 noro 1750: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1751: if ( c > 0 ) {
1.34 noro 1752: if ( sum ) HCM(gj) = sum;
1753: else g->body[j] = nd_remove_head(gj);
1.1 noro 1754: j = i;
1755: gj = g->body[j];
1756: dj = HDL(gj);
1.14 noro 1757: sum = HCM(gj);
1.1 noro 1758: } else if ( c == 0 ) {
1.19 noro 1759: sum = sum+HCM(gi)-mod;
1.34 noro 1760: if ( sum < 0 ) sum += mod;
1.1 noro 1761: g->body[i] = nd_remove_head(gi);
1762: }
1763: }
1764: }
1.34 noro 1765: if ( j < 0 ) return -1;
1.1 noro 1766: else if ( sum ) {
1.14 noro 1767: HCM(gj) = sum;
1.26 noro 1768: return j;
1.31 noro 1769: } else
1.26 noro 1770: g->body[j] = nd_remove_head(gj);
1771: }
1772: }
1773:
1774: int head_pbucket_q(PGeoBucket g)
1775: {
1776: int j,i,c,k,nv;
1777: Q sum,t;
1778: ND gi,gj;
1779:
1780: k = g->m;
1781: while ( 1 ) {
1782: j = -1;
1783: for ( i = 0; i <= k; i++ ) {
1.34 noro 1784: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1785: if ( j < 0 ) {
1786: j = i;
1787: gj = g->body[j];
1788: sum = HCQ(gj);
1789: } else {
1790: nv = NV(gi);
1.34 noro 1791: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1792: if ( c > 0 ) {
1.34 noro 1793: if ( sum ) HCQ(gj) = sum;
1794: else g->body[j] = nd_remove_head(gj);
1.26 noro 1795: j = i;
1796: gj = g->body[j];
1797: sum = HCQ(gj);
1798: } else if ( c == 0 ) {
1799: addq(sum,HCQ(gi),&t);
1800: sum = t;
1801: g->body[i] = nd_remove_head(gi);
1802: }
1803: }
1804: }
1.34 noro 1805: if ( j < 0 ) return -1;
1.26 noro 1806: else if ( sum ) {
1807: HCQ(gj) = sum;
1.1 noro 1808: return j;
1.31 noro 1809: } else
1.1 noro 1810: g->body[j] = nd_remove_head(gj);
1811: }
1812: }
1813:
1.25 noro 1814: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1815: {
1.31 noro 1816: int i;
1.1 noro 1817: ND r,t;
1818:
1819: r = 0;
1.28 noro 1820: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1821: r = nd_add(mod,r,g->body[i]);
1.28 noro 1822: g->body[i] = 0;
1823: }
1824: g->m = -1;
1.1 noro 1825: return r;
1826: }
1827:
1.27 noro 1828: /* return value = 0 => input is not a GB */
1829:
1830: NODE nd_gb(int m,int checkonly)
1.1 noro 1831: {
1832: int i,nh,sugar,stat;
1.23 noro 1833: NODE r,g,t;
1.1 noro 1834: ND_pairs d;
1835: ND_pairs l;
1836: ND h,nf;
1.63 noro 1837: NDV nfv;
1.1 noro 1838:
1.23 noro 1839: g = 0; d = 0;
1840: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1841: d = update_pairs(d,g,i);
1842: g = update_base(g,i);
1843: }
1844: sugar = 0;
1845: while ( d ) {
1846: again:
1847: l = nd_minp(d,&d);
1.14 noro 1848: if ( SG(l) != sugar ) {
1849: sugar = SG(l);
1.1 noro 1850: fprintf(asir_out,"%d",sugar);
1851: }
1.53 noro 1852: stat = nd_sp(m,0,l,&h);
1.1 noro 1853: if ( !stat ) {
1854: NEXT(l) = d; d = l;
1.20 noro 1855: d = nd_reconstruct(m,0,d);
1.1 noro 1856: goto again;
1857: }
1.41 noro 1858: #if USE_GEOBUCKET
1.53 noro 1859: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,&nf);
1.41 noro 1860: #else
1.53 noro 1861: stat = nd_nf(m,h,nd_ps,!Top,&nf);
1.41 noro 1862: #endif
1.1 noro 1863: if ( !stat ) {
1864: NEXT(l) = d; d = l;
1.20 noro 1865: d = nd_reconstruct(m,0,d);
1.1 noro 1866: goto again;
1867: } else if ( nf ) {
1.27 noro 1868: if ( checkonly ) return 0;
1.1 noro 1869: printf("+"); fflush(stdout);
1.63 noro 1870: nd_removecont(m,nf);
1871: nfv = ndtondv(m,nf); nd_free(nf);
1872: nh = ndv_newps(nfv,0);
1.1 noro 1873: d = update_pairs(d,g,nh);
1874: g = update_base(g,nh);
1875: FREENDP(l);
1876: } else {
1877: printf("."); fflush(stdout);
1878: FREENDP(l);
1879: }
1880: }
1.53 noro 1881: for ( t = g; t; t = NEXT(t) ) BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1882: return g;
1883: }
1884:
1.23 noro 1885: NODE nd_gb_trace(int m)
1.20 noro 1886: {
1887: int i,nh,sugar,stat;
1.23 noro 1888: NODE r,g,t;
1.20 noro 1889: ND_pairs d;
1890: ND_pairs l;
1891: ND h,nf,nfq;
1.63 noro 1892: NDV nfv,nfqv;
1.20 noro 1893:
1.23 noro 1894: g = 0; d = 0;
1895: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1896: d = update_pairs(d,g,i);
1897: g = update_base(g,i);
1898: }
1899: sugar = 0;
1900: while ( d ) {
1901: again:
1902: l = nd_minp(d,&d);
1903: if ( SG(l) != sugar ) {
1904: sugar = SG(l);
1905: fprintf(asir_out,"%d",sugar);
1906: }
1.53 noro 1907: stat = nd_sp(m,0,l,&h);
1.20 noro 1908: if ( !stat ) {
1909: NEXT(l) = d; d = l;
1910: d = nd_reconstruct(m,1,d);
1911: goto again;
1912: }
1.41 noro 1913: #if USE_GEOBUCKET
1.53 noro 1914: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1915: #else
1.53 noro 1916: stat = nd_nf(m,h,nd_ps,!Top,&nf);
1.41 noro 1917: #endif
1.20 noro 1918: if ( !stat ) {
1919: NEXT(l) = d; d = l;
1920: d = nd_reconstruct(m,1,d);
1921: goto again;
1922: } else if ( nf ) {
1923: /* overflow does not occur */
1.53 noro 1924: nd_sp(0,1,l,&h);
1925: nd_nf(0,h,nd_ps_trace,!Top,&nfq);
1.20 noro 1926: if ( nfq ) {
1.63 noro 1927: /* failure; m|HC(nfq) */
1928: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1929:
1.20 noro 1930: printf("+"); fflush(stdout);
1.63 noro 1931: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1932: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1933: nh = ndv_newps(nfv,nfqv);
1.20 noro 1934: d = update_pairs(d,g,nh);
1935: g = update_base(g,nh);
1936: } else {
1937: printf("*"); fflush(stdout);
1938: }
1939: } else {
1940: printf("."); fflush(stdout);
1941: }
1942: FREENDP(l);
1943: }
1.23 noro 1944: for ( t = g; t; t = NEXT(t) )
1.53 noro 1945: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1946: return g;
1947: }
1948:
1.23 noro 1949: int ndv_compare(NDV *p1,NDV *p2)
1950: {
1.34 noro 1951: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1952: }
1953:
1954: int ndv_compare_rev(NDV *p1,NDV *p2)
1955: {
1.34 noro 1956: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1957: }
1958:
1.61 noro 1959: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1960: {
1961: int i,j,n,stat;
1962: NDV *w,*ps;
1963: ND nf,g;
1964: NODE t,a0,a;
1.45 noro 1965: struct oBaseSet base;
1.61 noro 1966: UINT **bound;
1.23 noro 1967:
1968: for ( n = 0, t = f; t; t = NEXT(t), n++ );
1969: ps = (NDV *)ALLOCA(n*sizeof(NDV));
1.61 noro 1970: bound = (UINT **)ALLOCA(n*sizeof(UINT *));
1.54 noro 1971: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) ps[i] = (NDV)BDY(t);
1.23 noro 1972: qsort(ps,n,sizeof(NDV),(int (*)(const void *,const void *))ndv_compare);
1.54 noro 1973: for ( i = 0; i < n; i++ ) bound[i] = ndv_compute_bound(ps[i]);
1.45 noro 1974: base.ps = (NDV *)ALLOCA((n-1)*sizeof(NDV));
1.61 noro 1975: base.bound = (UINT **)ALLOCA((n-1)*sizeof(UINT *));
1.45 noro 1976: base.len = n-1;
1.50 noro 1977: i = 0;
1978: while ( i < n ) {
1.45 noro 1979: for ( j = 0; j < i; j++ ) {
1980: base.ps[j] = ps[j]; base.bound[j] = bound[j];
1981: }
1982: for ( j = i+1; j < n; j++ ) {
1983: base.ps[j-1] = ps[j]; base.bound[j-1] = bound[j];
1984: }
1.23 noro 1985: g = ndvtond(m,ps[i]);
1.45 noro 1986: stat = nd_nf_direct(m,g,&base,1,&nf);
1.50 noro 1987: if ( !stat )
1.23 noro 1988: nd_reconstruct_direct(m,ps,n);
1989: else if ( !nf ) {
1990: printf("."); fflush(stdout);
1991: ndv_free(ps[i]);
1.50 noro 1992: for ( j = i+1; j < n; j++ ) {
1993: ps[j-1] = ps[j]; bound[j-1] = bound[j];
1994: }
1.23 noro 1995: n--;
1.50 noro 1996: base.len = n-1;
1.23 noro 1997: } else {
1998: printf("."); fflush(stdout);
1999: ndv_free(ps[i]);
1.24 noro 2000: nd_removecont(m,nf);
1.23 noro 2001: ps[i] = ndtondv(m,nf);
1.45 noro 2002: bound[i] = ndv_compute_bound(ps[i]);
1.23 noro 2003: nd_free(nf);
1.50 noro 2004: i++;
1.23 noro 2005: }
2006: }
1.45 noro 2007: printf("\n");
1.23 noro 2008: for ( a0 = 0, i = 0; i < n; i++ ) {
2009: NEXTNODE(a0,a);
2010: BDY(a) = (pointer)ps[i];
2011: }
2012: NEXT(a) = 0;
2013: return a0;
2014: }
2015:
1.1 noro 2016: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
2017: {
2018: ND_pairs d1,nd,cur,head,prev,remove;
2019:
2020: if ( !g ) return d;
2021: d = crit_B(d,t);
2022: d1 = nd_newpairs(g,t);
2023: d1 = crit_M(d1);
2024: d1 = crit_F(d1);
1.55 noro 2025: if ( do_weyl )
2026: head = d1;
2027: else {
2028: prev = 0; cur = head = d1;
2029: while ( cur ) {
2030: if ( crit_2( cur->i1,cur->i2 ) ) {
2031: remove = cur;
2032: if ( !prev ) head = cur = NEXT(cur);
2033: else cur = NEXT(prev) = NEXT(cur);
2034: FREENDP(remove);
2035: } else {
2036: prev = cur; cur = NEXT(cur);
2037: }
1.1 noro 2038: }
2039: }
2040: if ( !d )
2041: return head;
2042: else {
2043: nd = d;
1.34 noro 2044: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 2045: NEXT(nd) = head;
2046: return d;
2047: }
2048: }
2049:
2050: ND_pairs nd_newpairs( NODE g, int t )
2051: {
2052: NODE h;
1.61 noro 2053: UINT *dl;
1.34 noro 2054: int ts,s;
1.1 noro 2055: ND_pairs r,r0;
2056:
1.20 noro 2057: dl = DL(nd_psh[t]);
1.34 noro 2058: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 2059: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
2060: NEXTND_pairs(r0,r);
2061: r->i1 = (int)BDY(h);
2062: r->i2 = t;
1.20 noro 2063: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 2064: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
2065: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 2066: }
2067: NEXT(r) = 0;
2068: return r0;
2069: }
2070:
2071: ND_pairs crit_B( ND_pairs d, int s )
2072: {
2073: ND_pairs cur,head,prev,remove;
1.61 noro 2074: UINT *t,*tl,*lcm;
1.1 noro 2075: int td,tdl;
2076:
2077: if ( !d ) return 0;
1.20 noro 2078: t = DL(nd_psh[s]);
1.1 noro 2079: prev = 0;
2080: head = cur = d;
1.61 noro 2081: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 2082: while ( cur ) {
2083: tl = cur->lcm;
2084: if ( ndl_reducible(tl,t)
1.20 noro 2085: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
2086: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 2087: remove = cur;
2088: if ( !prev ) {
2089: head = cur = NEXT(cur);
2090: } else {
2091: cur = NEXT(prev) = NEXT(cur);
2092: }
2093: FREENDP(remove);
2094: } else {
1.34 noro 2095: prev = cur; cur = NEXT(cur);
1.1 noro 2096: }
2097: }
2098: return head;
2099: }
2100:
2101: ND_pairs crit_M( ND_pairs d1 )
2102: {
2103: ND_pairs e,d2,d3,dd,p;
1.61 noro 2104: UINT *id,*jd;
1.1 noro 2105:
2106: for ( dd = 0, e = d1; e; e = d3 ) {
2107: if ( !(d2 = NEXT(e)) ) {
2108: NEXT(e) = dd;
2109: return e;
2110: }
1.34 noro 2111: id = LCM(e);
1.1 noro 2112: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2113: p = NEXT(d2);
2114: jd = LCM(d2);
2115: if ( ndl_equal(jd,id) )
2116: ;
2117: else if ( TD(jd) > TD(id) )
1.1 noro 2118: if ( ndl_reducible(jd,id) ) continue;
2119: else ;
1.34 noro 2120: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2121: NEXT(d2) = d3;
2122: d3 = d2;
2123: }
2124: NEXT(e) = dd;
2125: dd = e;
2126: continue;
2127: /**/
2128: delit: NEXT(d2) = d3;
2129: d3 = d2;
2130: for ( ; p; p = d2 ) {
2131: d2 = NEXT(p);
2132: NEXT(p) = d3;
2133: d3 = p;
2134: }
2135: FREENDP(e);
2136: }
2137: return dd;
2138: }
2139:
2140: ND_pairs crit_F( ND_pairs d1 )
2141: {
2142: ND_pairs rest, head,remove;
2143: ND_pairs last, p, r, w;
2144: int s;
2145:
2146: for ( head = last = 0, p = d1; NEXT(p); ) {
2147: r = w = equivalent_pairs(p,&rest);
1.14 noro 2148: s = SG(r);
1.1 noro 2149: w = NEXT(w);
2150: while ( w ) {
2151: if ( crit_2(w->i1,w->i2) ) {
2152: r = w;
2153: w = NEXT(w);
2154: while ( w ) {
2155: remove = w;
2156: w = NEXT(w);
2157: FREENDP(remove);
2158: }
2159: break;
1.14 noro 2160: } else if ( SG(w) < s ) {
1.1 noro 2161: FREENDP(r);
2162: r = w;
1.14 noro 2163: s = SG(r);
1.1 noro 2164: w = NEXT(w);
2165: } else {
2166: remove = w;
2167: w = NEXT(w);
2168: FREENDP(remove);
2169: }
2170: }
2171: if ( last ) NEXT(last) = r;
2172: else head = r;
2173: NEXT(last = r) = 0;
2174: p = rest;
2175: if ( !p ) return head;
2176: }
2177: if ( !last ) return p;
2178: NEXT(last) = p;
2179: return head;
2180: }
2181:
2182: int crit_2( int dp1, int dp2 )
2183: {
1.20 noro 2184: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2185: }
2186:
1.40 noro 2187: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2188: {
2189: ND_pairs w,p,r,s;
1.61 noro 2190: UINT *d;
1.1 noro 2191:
2192: w = d1;
1.34 noro 2193: d = LCM(w);
1.1 noro 2194: s = NEXT(w);
2195: NEXT(w) = 0;
2196: for ( r = 0; s; s = p ) {
2197: p = NEXT(s);
1.34 noro 2198: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2199: NEXT(s) = w; w = s;
1.1 noro 2200: } else {
1.39 noro 2201: NEXT(s) = r; r = s;
1.1 noro 2202: }
2203: }
2204: *prest = r;
2205: return w;
2206: }
2207:
2208: NODE update_base(NODE nd,int ndp)
2209: {
1.61 noro 2210: UINT *dl, *dln;
1.1 noro 2211: NODE last, p, head;
2212:
1.20 noro 2213: dl = DL(nd_psh[ndp]);
1.1 noro 2214: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2215: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2216: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2217: p = NEXT(p);
2218: if ( last ) NEXT(last) = p;
2219: } else {
2220: if ( !last ) head = p;
2221: p = NEXT(last = p);
2222: }
2223: }
2224: head = append_one(head,ndp);
2225: return head;
2226: }
2227:
2228: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2229: {
2230: ND_pairs m,ml,p,l;
1.61 noro 2231: UINT *lcm;
1.33 noro 2232: int s,td,len,tlen,c,c1;
1.1 noro 2233:
2234: if ( !(p = NEXT(m = d)) ) {
2235: *prest = p;
2236: NEXT(m) = 0;
2237: return m;
2238: }
1.14 noro 2239: s = SG(m);
1.33 noro 2240: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1.34 noro 2241: if ( (SG(p) < s)
2242: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
1.39 noro 2243: ml = l; m = p; s = SG(m);
1.1 noro 2244: }
2245: if ( !ml ) *prest = NEXT(m);
2246: else {
2247: NEXT(ml) = NEXT(m);
2248: *prest = d;
2249: }
2250: NEXT(m) = 0;
2251: return m;
2252: }
2253:
1.63 noro 2254: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2255: {
2256: int msugar;
2257: ND_pairs t,dm0,dm,dr0,dr;
2258:
2259: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2260: if ( SG(t) < msugar ) msugar = SG(t);
2261: dm0 = 0; dr0 = 0;
2262: for ( t = d; t; t = NEXT(t) )
2263: if ( SG(t) == msugar ) {
2264: if ( dm0 ) NEXT(dm) = t;
2265: else dm0 = t;
2266: dm = t;
2267: } else {
2268: if ( dr0 ) NEXT(dr) = t;
2269: else dr0 = t;
2270: dr = t;
2271: }
2272: NEXT(dm) = 0;
2273: if ( dr0 ) NEXT(dr) = 0;
2274: *prest = dr0;
2275: return dm0;
2276: }
2277:
2278: int ndv_newps(NDV a,NDV aq)
1.1 noro 2279: {
1.3 noro 2280: int len;
1.13 noro 2281: RHist r;
1.20 noro 2282: NDV b;
1.3 noro 2283:
1.1 noro 2284: if ( nd_psn == nd_pslen ) {
2285: nd_pslen *= 2;
1.11 noro 2286: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2287: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2288: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2289: nd_bound = (UINT **)
2290: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2291: }
1.39 noro 2292: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2293: nd_ps[nd_psn] = a;
1.39 noro 2294: if ( aq ) {
1.63 noro 2295: nd_ps_trace[nd_psn] = aq;
2296: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2297: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2298: } else {
1.63 noro 2299: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2300: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2301: }
1.1 noro 2302: return nd_psn++;
2303: }
2304:
1.61 noro 2305: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2306: {
1.5 noro 2307: int i,j,td,len,max;
1.1 noro 2308: NODE s,s0,f0;
1.61 noro 2309: UINT *d;
1.13 noro 2310: RHist r;
1.61 noro 2311: NDV a,am;
1.11 noro 2312:
2313: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2314:
2315: nd_psn = length(f); nd_pslen = 2*nd_psn;
1.11 noro 2316: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2317: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2318: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2319: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2320:
1.11 noro 2321: if ( !nd_red )
1.13 noro 2322: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
2323: bzero(nd_red,REDTAB_LEN*sizeof(RHist));
1.1 noro 2324: nd_free_private_storage();
2325: for ( i = 0; i < nd_psn; i++, f = NEXT(f) ) {
1.39 noro 2326: if ( trace ) {
1.61 noro 2327: a = nd_ps_trace[i] = ndv_dup(0,(NDV)BDY(f));
2328: ndv_removecont(0,a);
2329: am = nd_ps[i] = ndv_dup(mod,a);
2330: ndv_mod(mod,am);
2331: ndv_removecont(mod,am);
2332: } else {
2333: a = nd_ps[i] = ndv_dup(mod,(NDV)BDY(f));
2334: if ( mod ) ndv_mod(mod,a);
2335: ndv_removecont(mod,a);
1.39 noro 2336: }
1.61 noro 2337: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2338: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2339: nd_psh[i] = r;
2340: }
2341: }
2342:
1.63 noro 2343: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2344: {
1.61 noro 2345: VL tv,fv,vv,vc;
1.1 noro 2346: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2347: int e,max,nvar;
2348: ND b;
1.1 noro 2349:
2350: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.61 noro 2351: nvar = length(vv);
1.32 noro 2352: nd_init_ord(ord);
1.61 noro 2353: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2354: for ( tv = vv; tv; tv = NEXT(tv) ) {
2355: e = getdeg(tv->v,(P)BDY(t));
2356: max = MAX(e,max);
2357: }
2358: nd_setup_parameters(nvar,max);
1.1 noro 2359: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2360: b = (pointer)ptondv(CO,vv,(P)BDY(t));
2361: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2362: }
2363: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2364: ndv_setup(m,0,fd0);
1.63 noro 2365: x = f4?nd_f4(m):nd_gb(m,0);
1.61 noro 2366: x = ndv_reducebase(x);
2367: x = ndv_reduceall(m,x);
1.23 noro 2368: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2369: NEXTNODE(r0,r);
1.61 noro 2370: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2371: }
2372: if ( r0 ) NEXT(r) = 0;
2373: MKLIST(*rp,r0);
2374: }
2375:
1.52 noro 2376: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2377: {
2378: struct order_spec ord1;
1.61 noro 2379: VL tv,fv,vv,vc;
1.27 noro 2380: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2381: int m,nocheck,nvar,mindex,e,max;
2382: NDV c;
2383: NMV a;
1.27 noro 2384: P p;
1.61 noro 2385: EPOS oepos;
2386: int obpe,oadv,wmax,i,len,cbpe;
1.20 noro 2387:
2388: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.52 noro 2389: nvar = length(vv);
2390: nocheck = 0;
2391: mindex = 0;
2392:
2393: /* setup modulus */
2394: if ( trace < 0 ) {
2395: trace = -trace;
2396: nocheck = 1;
2397: }
2398: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2399: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2400: for ( tv = vv; tv; tv = NEXT(tv) ) {
2401: e = getdeg(tv->v,(P)BDY(t));
2402: max = MAX(e,max);
1.23 noro 2403: }
1.61 noro 2404: nd_init_ord(ord);
2405: nd_setup_parameters(nvar,max);
2406: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2407: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2408: c = ptondv(CO,vv,(P)BDY(t));
2409: if ( c ) {
2410: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2411: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2412: }
1.61 noro 2413: }
2414: if ( in0 ) NEXT(in) = 0;
2415: if ( fd0 ) NEXT(fd) = 0;
2416: if ( homo ) {
2417: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2418: c = (NDV)BDY(t); len = LEN(c);
2419: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2420: wmax = MAX(TD(DL(a)),wmax);
2421: }
2422: homogenize_order(ord,nvar,&ord1);
2423: nd_init_ord(&ord1);
2424: nd_setup_parameters(nvar+1,wmax);
2425: for ( t = fd0; t; t = NEXT(t) )
2426: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2427: }
1.52 noro 2428: while ( 1 ) {
1.61 noro 2429: ndv_setup(m,1,fd0);
1.27 noro 2430: cand = nd_gb_trace(m);
1.52 noro 2431: if ( !cand ) {
2432: /* failure */
1.61 noro 2433: if ( trace > 1 ) { *rp = 0; return; }
2434: else m = get_lprime(++mindex);
1.52 noro 2435: continue;
2436: }
1.27 noro 2437: if ( homo ) {
2438: /* dehomogenization */
1.61 noro 2439: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2440: nd_init_ord(ord);
1.61 noro 2441: nd_setup_parameters(nvar,0);
1.27 noro 2442: }
1.61 noro 2443: cand = ndv_reducebase(cand);
2444: cand = ndv_reduceall(0,cand);
2445: if ( nocheck )
2446: break;
2447: cbpe = nd_bpe;
2448: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2449: /* success */
2450: break;
2451: else if ( trace > 1 ) {
2452: /* failure */
2453: *rp = 0; return;
1.61 noro 2454: } else {
1.52 noro 2455: /* try the next modulus */
2456: m = get_lprime(++mindex);
1.61 noro 2457: /* reset the parameters */
2458: if ( homo ) {
2459: nd_init_ord(&ord1);
2460: nd_setup_parameters(nvar+1,wmax);
2461: } else {
2462: nd_init_ord(ord);
2463: nd_setup_parameters(nvar,max);
2464: }
2465: }
1.52 noro 2466: }
1.27 noro 2467: /* dp->p */
1.61 noro 2468: nd_bpe = cbpe;
2469: nd_setup_parameters(0,0);
2470: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2471: MKLIST(*rp,cand);
1.1 noro 2472: }
2473:
1.61 noro 2474: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2475: {
1.61 noro 2476: UINT *d;
1.57 noro 2477: int i,j,l,s,ord_l;
1.43 noro 2478: struct order_pair *op;
1.1 noro 2479:
2480: d = dl->d;
1.41 noro 2481: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2482: if ( nd_blockmask ) {
2483: l = nd_blockmask->n;
2484: op = nd_blockmask->order_pair;
2485: for ( j = 0, s = 0; j < l; j++ ) {
2486: ord_l = op[j].length;
1.57 noro 2487: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2488: }
2489: TD(r) = ndl_weight(r);
1.61 noro 2490: ndl_weight_mask(r);
1.43 noro 2491: } else {
1.56 noro 2492: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2493: TD(r) = ndl_weight(r);
2494: }
1.1 noro 2495: }
2496:
1.61 noro 2497: DL ndltodl(int n,UINT *ndl)
1.1 noro 2498: {
2499: DL dl;
2500: int *d;
1.57 noro 2501: int i,j,l,s,ord_l;
1.43 noro 2502: struct order_pair *op;
1.1 noro 2503:
2504: NEWDL(dl,n);
1.34 noro 2505: dl->td = TD(ndl);
1.1 noro 2506: d = dl->d;
1.43 noro 2507: if ( nd_blockmask ) {
2508: l = nd_blockmask->n;
2509: op = nd_blockmask->order_pair;
2510: for ( j = 0, s = 0; j < l; j++ ) {
2511: ord_l = op[j].length;
1.57 noro 2512: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2513: }
2514: } else {
1.56 noro 2515: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2516: }
1.1 noro 2517: return dl;
2518: }
2519:
1.61 noro 2520: void ndl_print(UINT *dl)
1.1 noro 2521: {
2522: int n;
1.57 noro 2523: int i,j,l,ord_l,s,s0;
1.43 noro 2524: struct order_pair *op;
1.1 noro 2525:
2526: n = nd_nvar;
2527: printf("<<");
1.43 noro 2528: if ( nd_blockmask ) {
2529: l = nd_blockmask->n;
2530: op = nd_blockmask->order_pair;
2531: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2532: ord_l = op[j].length;
1.57 noro 2533: for ( i = 0; i < ord_l; i++, s++ )
2534: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2535: }
2536: } else {
1.56 noro 2537: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2538: }
1.1 noro 2539: printf(">>");
2540: }
2541:
2542: void nd_print(ND p)
2543: {
2544: NM m;
2545:
2546: if ( !p )
2547: printf("0\n");
2548: else {
2549: for ( m = BDY(p); m; m = NEXT(m) ) {
1.14 noro 2550: printf("+%d*",CM(m));
2551: ndl_print(DL(m));
1.1 noro 2552: }
2553: printf("\n");
2554: }
2555: }
2556:
1.16 noro 2557: void nd_print_q(ND p)
2558: {
2559: NM m;
2560:
2561: if ( !p )
2562: printf("0\n");
2563: else {
2564: for ( m = BDY(p); m; m = NEXT(m) ) {
2565: printf("+");
2566: printexpr(CO,CQ(m));
2567: printf("*");
2568: ndl_print(DL(m));
2569: }
2570: printf("\n");
2571: }
2572: }
2573:
1.1 noro 2574: void ndp_print(ND_pairs d)
2575: {
2576: ND_pairs t;
2577:
1.34 noro 2578: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2579: printf("\n");
2580: }
2581:
1.20 noro 2582: void nd_removecont(int mod,ND p)
1.16 noro 2583: {
2584: int i,n;
2585: Q *w;
2586: Q dvr,t;
2587: NM m;
1.21 noro 2588: struct oVECT v;
2589: N q,r;
1.16 noro 2590:
1.34 noro 2591: if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2592: else {
2593: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2594: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2595: v.len = n;
2596: v.body = (pointer *)w;
1.34 noro 2597: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2598: removecont_array(w,n);
2599: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2600: }
2601: }
2602:
1.21 noro 2603: void nd_removecont2(ND p1,ND p2)
2604: {
2605: int i,n1,n2,n;
2606: Q *w;
2607: Q dvr,t;
2608: NM m;
2609: struct oVECT v;
2610: N q,r;
2611:
2612: if ( !p1 ) {
2613: nd_removecont(0,p2); return;
2614: } else if ( !p2 ) {
2615: nd_removecont(0,p1); return;
2616: }
2617: n1 = nd_length(p1);
2618: n2 = nd_length(p2);
2619: n = n1+n2;
2620: w = (Q *)ALLOCA(n*sizeof(Q));
2621: v.len = n;
2622: v.body = (pointer *)w;
1.34 noro 2623: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2624: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2625: removecont_array(w,n);
2626: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2627: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2628: }
2629:
1.20 noro 2630: void ndv_removecont(int mod,NDV p)
1.16 noro 2631: {
2632: int i,len;
2633: Q *w;
2634: Q dvr,t;
2635: NMV m;
2636:
1.20 noro 2637: if ( mod )
2638: ndv_mul_c(mod,p,invm(HCM(p),mod));
2639: else {
2640: len = p->len;
2641: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2642: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2643: sortbynm(w,len);
2644: qltozl(w,len,&dvr);
2645: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2646: divq(CQ(m),dvr,&t); CQ(m) = t;
2647: }
1.16 noro 2648: }
1.21 noro 2649: }
2650:
1.61 noro 2651: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2652: {
2653: int len,i,max;
2654: NMV m,mr0,mr,t;
2655:
2656: len = p->len;
2657: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2658: max = MAX(max,TD(DL(m)));
2659: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2660: m = (NMV)((char *)mr0+(len-1)*oadv);
2661: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2662: t = (NMV)ALLOCA(nmv_adv);
2663: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2664: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2665: CQ(mr) = CQ(m);
2666: ndl_copy(DL(t),DL(mr));
2667: }
2668: NV(p)++;
2669: BDY(p) = mr0;
2670: }
2671:
1.45 noro 2672: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2673: {
1.45 noro 2674: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2675: Q *w;
2676: Q dvr,t;
2677: NMV m,r;
2678:
2679: len = p->len;
2680: newnvar = nd_nvar-1;
1.48 noro 2681: newexporigin = nd_get_exporigin(ord);
1.45 noro 2682: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2683: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2684: ndl_dehomogenize(DL(m));
1.23 noro 2685: if ( newwpd != nd_wpd ) {
1.61 noro 2686: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2687: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2688: CQ(r) = CQ(m);
2689: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2690: adj = nd_exporigin-newexporigin;
2691: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2692: }
2693: }
2694: NV(p)--;
2695: }
2696:
1.21 noro 2697: void removecont_array(Q *c,int n)
2698: {
2699: struct oVECT v;
2700: Q d0,d1,a,u,u1,gcd;
2701: int i;
2702: N qn,rn,gn;
2703: Q *q,*r;
2704:
2705: q = (Q *)ALLOCA(n*sizeof(Q));
2706: r = (Q *)ALLOCA(n*sizeof(Q));
2707: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2708: igcdv_estimate(&v,&d0);
2709: for ( i = 0; i < n; i++ ) {
2710: divn(NM(c[i]),NM(d0),&qn,&rn);
2711: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2712: NTOQ(rn,SGN(c[i]),r[i]);
2713: }
1.34 noro 2714: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2715: if ( i < n ) {
2716: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2717: igcdv(&v,&d1);
2718: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2719: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2720: for ( i = 0; i < n; i++ ) {
2721: mulq(a,q[i],&u);
2722: if ( r[i] ) {
2723: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2724: addq(u,u1,&q[i]);
2725: } else
2726: q[i] = u;
2727: }
2728: }
1.34 noro 2729: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2730: }
2731:
1.19 noro 2732: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2733: {
2734: NM m;
2735: int c,c1;
2736:
1.34 noro 2737: if ( !p ) return;
1.1 noro 2738: for ( m = BDY(p); m; m = NEXT(m) ) {
1.14 noro 2739: c1 = CM(m);
1.19 noro 2740: DMAR(c1,mul,0,mod,c);
1.14 noro 2741: CM(m) = c;
1.1 noro 2742: }
2743: }
2744:
1.16 noro 2745: void nd_mul_c_q(ND p,Q mul)
2746: {
2747: NM m;
2748: Q c;
2749:
1.34 noro 2750: if ( !p ) return;
1.16 noro 2751: for ( m = BDY(p); m; m = NEXT(m) ) {
2752: mulq(CQ(m),mul,&c); CQ(m) = c;
2753: }
2754: }
2755:
1.61 noro 2756: void nd_mul_c_p(VL vl,ND p,P mul)
2757: {
2758: NM m;
2759: P c;
2760:
2761: if ( !p ) return;
2762: for ( m = BDY(p); m; m = NEXT(m) ) {
2763: mulp(vl,CP(m),mul,&c); CP(m) = c;
2764: }
2765: }
2766:
1.1 noro 2767: void nd_free(ND p)
2768: {
2769: NM t,s;
2770:
1.34 noro 2771: if ( !p ) return;
1.1 noro 2772: t = BDY(p);
2773: while ( t ) {
2774: s = NEXT(t);
2775: FREENM(t);
2776: t = s;
2777: }
2778: FREEND(p);
2779: }
2780:
1.23 noro 2781: void ndv_free(NDV p)
2782: {
2783: GC_free(BDY(p));
2784: }
2785:
1.61 noro 2786: void nd_append_red(UINT *d,int i)
1.1 noro 2787: {
1.13 noro 2788: RHist m,m0;
1.1 noro 2789: int h;
2790:
1.13 noro 2791: NEWRHist(m);
1.34 noro 2792: h = ndl_hash_value(d);
1.13 noro 2793: m->index = i;
1.14 noro 2794: ndl_copy(d,DL(m));
1.1 noro 2795: NEXT(m) = nd_red[h];
2796: nd_red[h] = m;
2797: }
2798:
1.61 noro 2799: UINT *ndv_compute_bound(NDV p)
1.1 noro 2800: {
1.61 noro 2801: UINT *d1,*d2,*t;
2802: UINT u;
1.57 noro 2803: int i,j,k,l,len,ind;
1.45 noro 2804: NMV m;
1.1 noro 2805:
2806: if ( !p )
2807: return 0;
1.61 noro 2808: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2809: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2810: len = LEN(p);
2811: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2812: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2813: ndl_lcm(DL(m),d1,d2);
1.1 noro 2814: t = d1; d1 = d2; d2 = t;
2815: }
1.12 noro 2816: l = nd_nvar+31;
1.61 noro 2817: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2818: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2819: u = d1[i];
2820: k = (nd_epw-1)*nd_bpe;
2821: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2822: t[ind] = (u>>k)&nd_mask0;
2823: }
2824: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2825: return t;
2826: }
2827:
1.48 noro 2828: int nd_get_exporigin(struct order_spec *ord)
2829: {
1.51 noro 2830: switch ( ord->id ) {
1.41 noro 2831: case 0:
1.48 noro 2832: return 1;
1.41 noro 2833: case 1:
2834: /* block order */
1.43 noro 2835: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2836: return ord->ord.block.length+1;
1.41 noro 2837: case 2:
1.52 noro 2838: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2839: }
1.48 noro 2840: }
2841:
1.61 noro 2842: void nd_setup_parameters(int nvar,int max) {
1.57 noro 2843: int i,j,n,elen,ord_o,ord_l,l,s;
2844: struct order_pair *op;
1.48 noro 2845:
1.61 noro 2846: /* if max == 0, don't touch nd_bpe */
2847: if ( max > 0 ) {
1.63 noro 2848: if ( max < 2 ) nd_bpe = 1;
1.62 noro 2849: if ( max < 4 ) nd_bpe = 2;
2850: else if ( max < 8 ) nd_bpe = 3;
2851: else if ( max < 16 ) nd_bpe = 4;
1.63 noro 2852: else if ( max < 32 ) nd_bpe = 5;
1.61 noro 2853: else if ( max < 64 ) nd_bpe = 6;
2854: else if ( max < 256 ) nd_bpe = 8;
1.63 noro 2855: else if ( max < 1024 ) nd_bpe = 10;
1.61 noro 2856: else if ( max < 65536 ) nd_bpe = 16;
2857: else nd_bpe = 32;
2858: }
2859: /* nvar == 0, don't touch nd_nvar */
2860: if ( nvar > 0 ) nd_nvar = nvar;
2861:
2862: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2863: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2864:
2865: nd_exporigin = nd_get_exporigin(nd_ord);
1.43 noro 2866: nd_wpd = nd_exporigin+elen;
1.57 noro 2867:
1.1 noro 2868: if ( nd_bpe < 32 ) {
2869: nd_mask0 = (1<<nd_bpe)-1;
2870: } else {
2871: nd_mask0 = 0xffffffff;
2872: }
2873: bzero(nd_mask,sizeof(nd_mask));
2874: nd_mask1 = 0;
2875: for ( i = 0; i < nd_epw; i++ ) {
2876: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2877: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2878: }
1.61 noro 2879: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2880: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2881: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2882: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2883: }
2884:
1.20 noro 2885: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2886: {
1.37 noro 2887: int i,obpe,oadv,h;
1.13 noro 2888: NM prev_nm_free_list;
2889: RHist mr0,mr;
2890: RHist r;
1.37 noro 2891: RHist *old_red;
1.1 noro 2892: ND_pairs s0,s,t,prev_ndp_free_list;
1.43 noro 2893: EPOS oepos;
1.15 noro 2894:
1.1 noro 2895: obpe = nd_bpe;
1.11 noro 2896: oadv = nmv_adv;
1.43 noro 2897: oepos = nd_epos;
1.63 noro 2898: if ( obpe < 2 ) nd_bpe = 2;
2899: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2900: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2901: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2902: else if ( obpe < 6 ) nd_bpe = 6;
2903: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2904: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2905: else if ( obpe < 16 ) nd_bpe = 16;
2906: else if ( obpe < 32 ) nd_bpe = 32;
2907: else error("nd_reconstruct : exponent too large");
1.5 noro 2908:
1.61 noro 2909: nd_setup_parameters(0,0);
1.1 noro 2910: prev_nm_free_list = _nm_free_list;
2911: prev_ndp_free_list = _ndp_free_list;
2912: _nm_free_list = 0;
2913: _ndp_free_list = 0;
1.53 noro 2914: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2915: if ( trace )
2916: for ( i = nd_psn-1; i >= 0; i-- )
2917: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2918: s0 = 0;
2919: for ( t = d; t; t = NEXT(t) ) {
2920: NEXTND_pairs(s0,s);
2921: s->i1 = t->i1;
2922: s->i2 = t->i2;
1.14 noro 2923: SG(s) = SG(t);
1.61 noro 2924: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2925: }
1.37 noro 2926:
2927: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2928: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2929: old_red[i] = nd_red[i];
2930: nd_red[i] = 0;
2931: }
2932: for ( i = 0; i < REDTAB_LEN; i++ )
2933: for ( r = old_red[i]; r; r = NEXT(r) ) {
2934: NEWRHist(mr);
1.13 noro 2935: mr->index = r->index;
1.20 noro 2936: SG(mr) = SG(r);
1.61 noro 2937: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2938: h = ndl_hash_value(DL(mr));
2939: NEXT(mr) = nd_red[h];
2940: nd_red[h] = mr;
1.6 noro 2941: }
1.37 noro 2942: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2943: old_red = 0;
1.11 noro 2944: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2945: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2946: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2947: nd_psh[i] = r;
1.11 noro 2948: }
1.1 noro 2949: if ( s0 ) NEXT(s) = 0;
2950: prev_nm_free_list = 0;
2951: prev_ndp_free_list = 0;
2952: GC_gcollect();
2953: return s0;
2954: }
2955:
1.23 noro 2956: void nd_reconstruct_direct(int mod,NDV *ps,int len)
2957: {
1.37 noro 2958: int i,obpe,oadv,h;
1.61 noro 2959: UINT **bound;
1.23 noro 2960: NM prev_nm_free_list;
2961: RHist mr0,mr;
2962: RHist r;
1.37 noro 2963: RHist *old_red;
1.23 noro 2964: ND_pairs s0,s,t,prev_ndp_free_list;
1.43 noro 2965: EPOS oepos;
1.23 noro 2966:
2967: obpe = nd_bpe;
2968: oadv = nmv_adv;
1.43 noro 2969: oepos = nd_epos;
1.63 noro 2970: if ( obpe < 2 ) nd_bpe = 2;
2971: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2972: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2973: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2974: else if ( obpe < 6 ) nd_bpe = 6;
2975: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2976: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2977: else if ( obpe < 16 ) nd_bpe = 16;
2978: else if ( obpe < 32 ) nd_bpe = 32;
2979: else error("nd_reconstruct_direct : exponent too large");
1.23 noro 2980:
1.61 noro 2981: nd_setup_parameters(0,0);
1.23 noro 2982: prev_nm_free_list = _nm_free_list;
2983: prev_ndp_free_list = _ndp_free_list;
1.34 noro 2984: _nm_free_list = 0; _ndp_free_list = 0;
1.43 noro 2985: for ( i = len-1; i >= 0; i-- ) ndv_realloc(ps[i],obpe,oadv,oepos);
1.23 noro 2986: prev_nm_free_list = 0;
2987: prev_ndp_free_list = 0;
2988: GC_gcollect();
2989: }
2990:
1.61 noro 2991: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2992: {
1.57 noro 2993: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2994: struct order_pair *op;
1.1 noro 2995:
2996: n = nd_nvar;
1.61 noro 2997: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2998: omask0 = (1<<obpe)-1;
1.34 noro 2999: TD(r) = TD(d);
1.41 noro 3000: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3001: if ( nd_blockmask ) {
3002: l = nd_blockmask->n;
3003: op = nd_blockmask->order_pair;
3004: for ( i = 1; i < nd_exporigin; i++ )
3005: r[i] = d[i];
3006: for ( j = 0, s = 0; j < l; j++ ) {
3007: ord_l = op[j].length;
1.57 noro 3008: for ( i = 0; i < ord_l; i++, s++ ) {
3009: ei = GET_EXP_OLD(d,s);
3010: PUT_EXP(r,s,ei);
3011: }
1.1 noro 3012: }
1.43 noro 3013: } else {
1.56 noro 3014: for ( i = 0; i < n; i++ ) {
3015: ei = GET_EXP_OLD(d,i);
3016: PUT_EXP(r,i,ei);
3017: }
1.1 noro 3018: }
3019: }
1.3 noro 3020:
1.6 noro 3021: ND nd_copy(ND p)
3022: {
3023: NM m,mr,mr0;
1.41 noro 3024: int c,n;
1.6 noro 3025: ND r;
3026:
3027: if ( !p )
3028: return 0;
3029: else {
3030: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3031: NEXTNM(mr0,mr);
1.14 noro 3032: CM(mr) = CM(m);
3033: ndl_copy(DL(m),DL(mr));
1.6 noro 3034: }
3035: NEXT(mr) = 0;
1.31 noro 3036: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3037: SG(r) = SG(p);
1.6 noro 3038: return r;
3039: }
3040: }
3041:
1.53 noro 3042: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3043: {
3044: NM m;
3045: NDV p1,p2;
3046: ND t1,t2;
1.61 noro 3047: UINT *lcm;
1.31 noro 3048: int td;
1.11 noro 3049:
1.53 noro 3050: if ( trace ) {
3051: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3052: } else {
1.20 noro 3053: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3054: }
1.34 noro 3055: lcm = LCM(p);
1.11 noro 3056: NEWNM(m);
1.20 noro 3057: CQ(m) = HCQ(p2);
1.34 noro 3058: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3059: if ( ndl_check_bound2(p->i1,DL(m)) )
3060: return 0;
1.55 noro 3061: t1 = ndv_mul_nm(mod,m,p1);
1.34 noro 3062: if ( mod ) CM(m) = mod-HCM(p1);
3063: else chsgnq(HCQ(p1),&CQ(m));
3064: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3065: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3066: nd_free(t1);
3067: return 0;
3068: }
1.55 noro 3069: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3070: *rp = nd_add(mod,t1,t2);
1.11 noro 3071: FREENM(m);
3072: return 1;
3073: }
3074:
1.19 noro 3075: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3076: {
3077: NMV m;
3078: int c,c1,len,i;
3079:
1.34 noro 3080: if ( !p ) return;
1.14 noro 3081: len = LEN(p);
1.11 noro 3082: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.34 noro 3083: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
1.11 noro 3084: }
3085: }
3086:
1.16 noro 3087: void ndv_mul_c_q(NDV p,Q mul)
3088: {
3089: NMV m;
3090: Q c;
3091: int len,i;
3092:
1.34 noro 3093: if ( !p ) return;
1.16 noro 3094: len = LEN(p);
3095: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3096: mulq(CQ(m),mul,&c); CQ(m) = c;
3097: }
3098: }
3099:
1.55 noro 3100: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3101: int n2,i,j,l,n,tlen;
1.61 noro 3102: UINT *d0;
1.55 noro 3103: NM *tab,*psum;
3104: ND s,r;
3105: NM t;
3106: NMV m1;
3107:
3108: if ( !p ) return 0;
3109: n = NV(p); n2 = n>>1;
3110: d0 = DL(m0);
3111: l = LEN(p);
3112: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3113: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3114: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3115: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3116: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3117: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3118: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3119: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3120: for ( j = 0; j < tlen; j++ ) {
3121: if ( tab[j] ) {
3122: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3123: }
3124: }
3125: }
3126: for ( i = tlen-1, r = 0; i >= 0; i-- )
3127: if ( psum[i] ) {
3128: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3129: MKND(n,psum[i],j,s);
3130: r = nd_add(mod,r,s);
3131: }
1.56 noro 3132: if ( r ) SG(r) = SG(p)+TD(d0);
3133: return r;
1.55 noro 3134: }
3135:
1.56 noro 3136: /* product of monomials */
3137: /* XXX block order is not handled correctly */
3138:
1.55 noro 3139: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3140: {
1.56 noro 3141: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3142: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3143: Q *ctab_q;
3144: Q q,q1;
1.61 noro 3145: UINT c0,c1,c;
1.55 noro 3146: NM *p;
3147: NM m,t;
3148:
3149: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3150: if ( !m0 || !m1 ) return;
3151: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3152: NEWNM(m); d = DL(m);
1.56 noro 3153: if ( mod ) {
3154: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3155: } else
3156: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3157: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3158: homo = n&1 ? 1 : 0;
3159: if ( homo ) {
3160: /* offset of h-degree */
3161: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3162: PUT_EXP(DL(m),n-1,h);
3163: TD(DL(m)) = h;
1.61 noro 3164: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3165: }
3166: tab[0] = m;
3167: NEWNM(m); d = DL(m);
1.57 noro 3168: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3169: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3170: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3171: /* xi^a*(Di^k*xi^l)*Di^b */
3172: a += l; b += k;
1.56 noro 3173: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3174: if ( !k || !l ) {
3175: for ( j = 0; j < curlen; j++ )
1.56 noro 3176: if ( t = tab[j] ) {
3177: dt = DL(t);
3178: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3179: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3180: }
3181: curlen *= k+1;
3182: continue;
3183: }
3184: min = MIN(k,l);
1.56 noro 3185: if ( mod ) {
1.61 noro 3186: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3187: mkwcm(k,l,mod,ctab);
3188: } else {
3189: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3190: mkwc(k,l,ctab_q);
3191: }
1.57 noro 3192: for ( j = min; j >= 0; j-- ) {
1.56 noro 3193: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3194: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3195: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3196: if ( homo ) {
3197: TD(d) = s;
1.56 noro 3198: PUT_EXP(d,n-1,s-h);
1.55 noro 3199: } else TD(d) = h;
1.61 noro 3200: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3201: if ( mod ) c = ctab[j];
3202: else q = ctab_q[j];
1.57 noro 3203: p = tab+curlen*j;
3204: if ( j == 0 ) {
3205: for ( u = 0; u < curlen; u++, p++ ) {
3206: if ( tab[u] ) {
3207: ndl_addto(DL(tab[u]),d);
3208: if ( mod ) {
3209: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3210: } else {
3211: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3212: }
3213: }
1.56 noro 3214: }
1.57 noro 3215: } else {
3216: for ( u = 0; u < curlen; u++, p++ ) {
3217: if ( tab[u] ) {
3218: NEWNM(t);
3219: ndl_add(DL(tab[u]),d,DL(t));
3220: if ( mod ) {
3221: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3222: } else
3223: mulq(CQ(tab[u]),q,&CQ(t));
3224: *p = t;
3225: }
1.55 noro 3226: }
3227: }
3228: }
3229: curlen *= k+1;
3230: }
3231: FREENM(m);
3232: }
3233:
1.63 noro 3234: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3235: {
3236: NM mr,mr0;
3237: NMV m;
3238: UINT *d,*dt,*dm;
3239: int c,n,td,i,c1,c2,len;
3240: Q q;
3241: ND r;
3242:
3243: if ( !p ) return 0;
3244: else {
3245: n = NV(p); m = BDY(p);
3246: d = DL(m0);
3247: len = LEN(p);
3248: mr0 = 0;
3249: td = TD(d);
3250: c = CM(m0);
3251: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3252: NEXTNM(mr0,mr);
3253: CM(mr) = 1;
3254: ndl_add(DL(m),d,DL(mr));
3255: }
3256: NEXT(mr) = 0;
3257: MKND(NV(p),mr0,len,r);
3258: SG(r) = SG(p) + TD(d);
3259: return r;
3260: }
3261: }
3262:
1.55 noro 3263: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3264: {
3265: NM mr,mr0;
3266: NMV m;
1.61 noro 3267: UINT *d,*dt,*dm;
1.9 noro 3268: int c,n,td,i,c1,c2,len;
1.16 noro 3269: Q q;
1.9 noro 3270: ND r;
3271:
1.34 noro 3272: if ( !p ) return 0;
1.55 noro 3273: else if ( do_weyl )
3274: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3275: else {
3276: n = NV(p); m = BDY(p);
1.34 noro 3277: d = DL(m0);
1.14 noro 3278: len = LEN(p);
1.9 noro 3279: mr0 = 0;
1.34 noro 3280: td = TD(d);
1.16 noro 3281: if ( mod ) {
3282: c = CM(m0);
3283: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3284: NEXTNM(mr0,mr);
3285: c1 = CM(m);
1.19 noro 3286: DMAR(c1,c,0,mod,c2);
1.16 noro 3287: CM(mr) = c2;
3288: ndl_add(DL(m),d,DL(mr));
3289: }
3290: } else {
3291: q = CQ(m0);
3292: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3293: NEXTNM(mr0,mr);
3294: mulq(CQ(m),q,&CQ(mr));
3295: ndl_add(DL(m),d,DL(mr));
3296: }
1.4 noro 3297: }
1.9 noro 3298: NEXT(mr) = 0;
1.31 noro 3299: MKND(NV(p),mr0,len,r);
1.34 noro 3300: SG(r) = SG(p) + TD(d);
1.9 noro 3301: return r;
1.4 noro 3302: }
3303: }
3304:
1.43 noro 3305: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3306: {
1.13 noro 3307: NMV m,mr,mr0,t;
3308: int len,i,k;
1.11 noro 3309:
1.61 noro 3310: if ( !p ) return;
3311: m = BDY(p); len = LEN(p);
3312: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3313: m = (NMV)((char *)mr0+(len-1)*oadv);
3314: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3315: t = (NMV)ALLOCA(nmv_adv);
3316: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3317: CQ(t) = CQ(m);
3318: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3319: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3320: CQ(mr) = CQ(t);
3321: ndl_copy(DL(t),DL(mr));
3322: }
3323: BDY(p) = mr0;
3324: }
3325:
3326: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3327: {
3328: NMV m,mr,mr0;
3329: int len,i;
3330: NDV r;
1.11 noro 3331:
1.61 noro 3332: if ( !p ) return 0;
3333: m = BDY(p); len = LEN(p);
3334: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3335: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3336: ndl_zero(DL(mr));
3337: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3338: CQ(mr) = CQ(m);
1.11 noro 3339: }
1.61 noro 3340: MKNDV(NV(p),mr0,len,r);
3341: SG(r) = SG(p);
3342: return r;
1.11 noro 3343: }
3344:
1.61 noro 3345: /* duplicate p */
3346:
3347: NDV ndv_dup(int mod,NDV p)
1.3 noro 3348: {
3349: NDV d;
1.61 noro 3350: NMV t,m,m0;
1.3 noro 3351: int i,len;
3352:
1.34 noro 3353: if ( !p ) return 0;
1.31 noro 3354: len = LEN(p);
1.34 noro 3355: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3356: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3357: ndl_copy(DL(t),DL(m));
1.16 noro 3358: CQ(m) = CQ(t);
1.3 noro 3359: }
3360: MKNDV(NV(p),m0,len,d);
1.23 noro 3361: SG(d) = SG(p);
3362: return d;
3363: }
3364:
1.63 noro 3365: ND nd_dup(ND p)
3366: {
3367: ND d;
3368: NM t,m,m0;
3369:
3370: if ( !p ) return 0;
3371: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3372: NEXTNM(m0,m);
3373: ndl_copy(DL(t),DL(m));
3374: CQ(m) = CQ(t);
3375: }
3376: if ( m0 ) NEXT(m) = 0;
3377: MKND(NV(p),m0,LEN(p),d);
3378: SG(d) = SG(p);
3379: return d;
3380: }
3381:
1.61 noro 3382: /* XXX if p->len == 0 then it represents 0 */
3383:
3384: void ndv_mod(int mod,NDV p)
3385: {
3386: NMV t,d;
3387: int r;
3388: int i,len,dlen;
3389:
3390: if ( !p ) return;
3391: len = LEN(p);
3392: dlen = 0;
3393: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3394: r = rem(NM(CQ(t)),mod);
3395: if ( r ) {
3396: if ( SGN(CQ(t)) < 0 )
3397: r = mod-r;
3398: CM(d) = r;
3399: ndl_copy(DL(t),DL(d));
3400: NMV_ADV(d);
3401: dlen++;
3402: }
3403: }
3404: LEN(p) = dlen;
3405: }
3406:
3407: NDV ptondv(VL vl,VL dvl,P p)
3408: {
3409: ND nd;
3410:
3411: nd = ptond(vl,dvl,p);
3412: return ndtondv(0,nd);
3413: }
3414:
3415: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3416: {
1.61 noro 3417: int n,i,j,k,e;
3418: VL tvl;
3419: V v;
3420: DCP dc;
3421: DCP *w;
3422: ND r,s,t,u;
3423: P x;
3424: int c;
3425: UINT *d;
1.23 noro 3426: NM m,m0;
1.61 noro 3427:
3428: if ( !p )
3429: return 0;
3430: else if ( NUM(p) ) {
3431: NEWNM(m);
3432: ndl_zero(DL(m));
3433: CQ(m) = (Q)p;
3434: NEXT(m) = 0;
3435: MKND(nd_nvar,m,1,r);
3436: SG(r) = 0;
3437: return r;
3438: } else {
3439: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3440: w = (DCP *)ALLOCA(k*sizeof(DCP));
3441: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3442: for ( i = 0, tvl = dvl, v = VR(p);
3443: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3444: if ( !tvl ) {
3445: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3446: t = ptond(vl,dvl,COEF(w[j]));
3447: pwrp(vl,x,DEG(w[j]),&p);
3448: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3449: }
3450: return s;
3451: } else {
3452: NEWNM(m0); d = DL(m0);
3453: for ( j = k-1, s = 0; j >= 0; j-- ) {
3454: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3455: TD(d) = MUL_WEIGHT(e,i);
3456: if ( nd_blockmask) ndl_weight_mask(d);
3457: t = ptond(vl,dvl,COEF(w[j]));
3458: for ( m = BDY(t); m; m = NEXT(m) )
3459: ndl_addto(DL(m),d);
3460: SG(t) += TD(d);
3461: s = nd_add(0,s,t);
3462: }
3463: FREENM(m0);
3464: return s;
3465: }
3466: }
3467: }
3468:
3469: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3470: {
3471: VL tvl;
3472: int len,n,j,i,e;
3473: NMV m;
3474: Q q;
3475: P c;
3476: UINT *d;
3477: P s,r,u,t,w;
1.23 noro 3478:
1.34 noro 3479: if ( !p ) return 0;
1.61 noro 3480: else {
3481: len = LEN(p);
3482: n = NV(p);
3483: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3484: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
3485: if ( mod ) {
3486: STOQ(CM(m),q); c = (P)q;
3487: } else
3488: c = CP(m);
3489: d = DL(m);
3490: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3491: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3492: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3493: }
3494: addp(vl,s,t,&u); s = u;
3495: }
3496: return s;
1.23 noro 3497: }
1.3 noro 3498: }
3499:
1.61 noro 3500: NDV ndtondv(int mod,ND p)
1.11 noro 3501: {
3502: NDV d;
1.61 noro 3503: NMV m,m0;
3504: NM t;
3505: int i,len;
1.11 noro 3506:
1.34 noro 3507: if ( !p ) return 0;
1.61 noro 3508: len = LEN(p);
3509: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
3510: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3511: ndl_copy(DL(t),DL(m));
3512: CQ(m) = CQ(t);
1.11 noro 3513: }
1.61 noro 3514: MKNDV(NV(p),m0,len,d);
1.14 noro 3515: SG(d) = SG(p);
1.11 noro 3516: return d;
3517: }
3518:
1.61 noro 3519: ND ndvtond(int mod,NDV p)
1.11 noro 3520: {
1.61 noro 3521: ND d;
3522: NM m,m0;
1.11 noro 3523: NMV t;
1.61 noro 3524: int i,len;
1.11 noro 3525:
1.34 noro 3526: if ( !p ) return 0;
1.11 noro 3527: m0 = 0;
1.61 noro 3528: len = p->len;
3529: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3530: NEXTNM(m0,m);
3531: ndl_copy(DL(t),DL(m));
3532: CQ(m) = CQ(t);
1.11 noro 3533: }
3534: NEXT(m) = 0;
1.61 noro 3535: MKND(NV(p),m0,len,d);
1.14 noro 3536: SG(d) = SG(p);
1.11 noro 3537: return d;
3538: }
3539:
1.3 noro 3540: void ndv_print(NDV p)
3541: {
3542: NMV m;
3543: int i,len;
3544:
1.34 noro 3545: if ( !p ) printf("0\n");
1.3 noro 3546: else {
1.14 noro 3547: len = LEN(p);
1.3 noro 3548: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.14 noro 3549: printf("+%d*",CM(m));
1.16 noro 3550: ndl_print(DL(m));
3551: }
3552: printf("\n");
3553: }
3554: }
3555:
3556: void ndv_print_q(NDV p)
3557: {
3558: NMV m;
3559: int i,len;
3560:
1.34 noro 3561: if ( !p ) printf("0\n");
1.16 noro 3562: else {
3563: len = LEN(p);
3564: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3565: printf("+");
3566: printexpr(CO,CQ(m));
3567: printf("*");
1.14 noro 3568: ndl_print(DL(m));
1.3 noro 3569: }
3570: printf("\n");
3571: }
1.25 noro 3572: }
3573:
1.61 noro 3574: NODE ndv_reducebase(NODE x)
1.27 noro 3575: {
3576: int len,i,j;
3577: NDV *w;
3578: NODE t,t0;
3579:
3580: len = length(x);
3581: w = (NDV *)ALLOCA(len*sizeof(NDV));
3582: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3583: for ( i = 0; i < len; i++ ) {
3584: for ( j = 0; j < i; j++ ) {
3585: if ( w[i] && w[j] )
3586: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3587: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3588: }
3589: }
3590: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3591: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3592: }
3593: NEXT(t) = 0; x = t0;
3594: return x;
1.11 noro 3595: }
1.32 noro 3596:
1.43 noro 3597: /* XXX incomplete */
3598:
1.32 noro 3599: void nd_init_ord(struct order_spec *ord)
3600: {
1.43 noro 3601: switch ( ord->id ) {
1.32 noro 3602: case 0:
1.43 noro 3603: switch ( ord->ord.simple ) {
3604: case 0:
3605: nd_dcomp = 1;
3606: nd_isrlex = 1;
3607: break;
3608: case 1:
3609: nd_dcomp = 1;
3610: nd_isrlex = 0;
3611: break;
3612: case 2:
3613: nd_dcomp = 0;
3614: nd_isrlex = 0;
1.45 noro 3615: ndl_compare_function = ndl_lex_compare;
1.58 noro 3616: break;
3617: case 11:
3618: /* XXX */
3619: nd_dcomp = 0;
3620: nd_isrlex = 1;
3621: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3622: break;
3623: default:
3624: error("nd_gr : unsupported order");
3625: }
1.32 noro 3626: break;
3627: case 1:
1.43 noro 3628: /* XXX */
3629: nd_dcomp = -1;
1.32 noro 3630: nd_isrlex = 0;
1.45 noro 3631: ndl_compare_function = ndl_block_compare;
1.34 noro 3632: break;
1.43 noro 3633: case 2:
3634: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3635: break;
3636: }
1.41 noro 3637: nd_ord = ord;
1.32 noro 3638: }
3639:
1.43 noro 3640: BlockMask nd_create_blockmask(struct order_spec *ord)
3641: {
3642: int n,i,j,s,l;
1.61 noro 3643: UINT *t;
1.43 noro 3644: BlockMask bm;
3645:
3646: if ( !ord->id )
3647: return 0;
3648: n = ord->ord.block.length;
3649: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3650: bm->n = n;
3651: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3652: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3653: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3654: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3655: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3656: l = bm->order_pair[i].length;
3657: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3658: }
3659: return bm;
1.57 noro 3660: }
3661:
3662: EPOS nd_create_epos(struct order_spec *ord)
3663: {
3664: int i,j,l,s,ord_l,ord_o;
3665: EPOS epos;
3666: struct order_pair *op;
3667:
3668: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3669: switch ( ord->id ) {
3670: case 0:
3671: if ( nd_isrlex ) {
3672: for ( i = 0; i < nd_nvar; i++ ) {
3673: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3674: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3675: }
3676: } else {
3677: for ( i = 0; i < nd_nvar; i++ ) {
3678: epos[i].i = nd_exporigin + i/nd_epw;
3679: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3680: }
3681: }
3682: break;
3683: case 1:
3684: /* block order */
3685: l = ord->ord.block.length;
3686: op = ord->ord.block.order_pair;
3687: for ( j = 0, s = 0; j < l; j++ ) {
3688: ord_o = op[j].order;
3689: ord_l = op[j].length;
3690: if ( !ord_o )
3691: for ( i = 0; i < ord_l; i++ ) {
3692: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3693: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3694: }
3695: else
3696: for ( i = 0; i < ord_l; i++ ) {
3697: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3698: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3699: }
3700: s += ord_l;
3701: }
3702: break;
3703: case 2:
3704: error("nd_create_epos : matrix order is not supported yet.");
3705: }
3706: return epos;
1.43 noro 3707: }
1.59 noro 3708:
3709: /* external interface */
3710:
3711: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3712: {
1.61 noro 3713: NODE t,in0,in;
1.59 noro 3714: ND nd,nf;
1.61 noro 3715: NDV ndv;
3716: VL vv,tv;
3717: int stat,nvar,max,e;
1.59 noro 3718:
3719: pltovl(v,&vv);
1.61 noro 3720: nvar = length(vv);
3721:
3722: /* get the degree bound */
3723: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3724: for ( tv = vv; tv; tv = NEXT(tv) ) {
3725: e = getdeg(tv->v,(P)BDY(t));
3726: max = MAX(e,max);
3727: }
3728: for ( tv = vv; tv; tv = NEXT(tv) ) {
3729: e = getdeg(tv->v,f);
3730: max = MAX(e,max);
3731: }
3732:
1.59 noro 3733: nd_init_ord(ord);
1.61 noro 3734: nd_setup_parameters(nvar,max);
3735:
3736: /* conversion to ndv */
3737: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3738: NEXTNODE(in0,in);
3739: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
3740: }
3741: NEXTNODE(in0,in);
3742: BDY(in) = (pointer)ptondv(CO,vv,f);
3743: NEXT(in) = 0;
3744:
3745: ndv_setup(m,0,in0);
1.59 noro 3746: nd_psn--;
3747: nd_scale=2;
3748: while ( 1 ) {
3749: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
3750: stat = nd_nf(m,nd,nd_ps,1,&nf);
3751: if ( !stat ) {
3752: nd_psn++;
3753: nd_reconstruct(m,0,0);
3754: nd_psn--;
3755: } else
3756: break;
3757: }
1.61 noro 3758: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3759: }
3760:
3761: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3762: {
3763: NM m;
3764: UINT *t,*s;
3765: int i;
3766:
3767: for ( i = 0; i < n; i++ ) r[i] = 0;
3768: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3769: t = DL(m);
3770: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3771: r[i] = CM(m);
3772: }
3773: for ( i = 0; !r[i]; i++ );
3774: return i;
3775: }
3776:
3777: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3778: {
3779: NM m;
3780: NMV mr;
3781: UINT *d,*t,*s;
3782: NDV p;
3783: int i,j,len;
3784:
3785: m = pair->mul;
3786: d = DL(m);
3787: p = nd_ps[pair->index];
3788: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3789: for ( i = 0; i < n; i++ ) r[i] = 0;
3790: len = LEN(p);
3791: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3792: ndl_add(d,DL(mr),t);
3793: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3794: r[i] = CM(mr);
3795: }
3796: for ( i = 0; !r[i]; i++ );
3797: return i;
3798: }
3799:
1.67 noro 3800: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3801: {
3802: NM m;
3803: NMV mr;
3804: UINT *d,*t,*s;
3805: NDV p;
1.67 noro 3806: unsigned char *ivc;
3807: unsigned short *ivs;
3808: UINT *v,*ivi;
3809: int i,j,len,prev,diff,cdiff;
3810: IndArray r;
1.64 noro 3811:
3812: m = pair->mul;
3813: d = DL(m);
3814: p = nd_ps[pair->index];
3815: len = LEN(p);
3816: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3817: r = (IndArray)MALLOC(sizeof(struct oIndArray));
3818: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3819: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3820: ndl_add(d,DL(mr),t);
3821: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3822: v[j] = i;
3823: }
3824: r->head = v[0];
3825: diff = 0;
3826: for ( i = 1; i < len; i++ ) {
3827: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3828: }
3829: if ( diff < 256 ) {
3830: r->width = 1;
3831: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3832: r->index.c = ivc;
3833: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3834: } else if ( diff < 65536 ) {
3835: r->width = 2;
3836: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3837: r->index.s = ivs;
3838: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3839: } else {
3840: r->width = 4;
3841: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3842: r->index.i = ivi;
3843: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3844: }
1.67 noro 3845: return r;
1.64 noro 3846: }
3847:
1.65 noro 3848:
1.67 noro 3849: void ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NODE rp0)
1.65 noro 3850: {
1.67 noro 3851: int i,j,k,len,pos,prev;
1.66 noro 3852: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3853: IndArray ivect;
3854: unsigned char *ivc;
3855: unsigned short *ivs;
3856: unsigned int *ivi;
1.65 noro 3857: NDV redv;
1.67 noro 3858: NMV mr;
1.65 noro 3859: NODE rp;
3860:
3861: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
3862: ivect = imat[i];
1.67 noro 3863: k = ivect->head; svect[k] %= m;
1.65 noro 3864: if ( c = svect[k] ) {
1.67 noro 3865: c = m-c; redv = nd_ps[((NM_ind_pair)BDY(rp))->index];
3866: len = LEN(redv); mr = BDY(redv);
3867: svect[k] = 0; prev = k;
3868: switch ( ivect->width ) {
3869: case 1:
3870: ivc = ivect->index.c;
3871: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3872: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3873: prev = pos;
3874: DMA(c1,c,c2,up,lo);
3875: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3876: } else svect[pos] = lo;
3877: }
3878: break;
3879: case 2:
3880: ivs = ivect->index.s;
3881: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3882: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3883: prev = pos;
3884: DMA(c1,c,c2,up,lo);
3885: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3886: } else svect[pos] = lo;
3887: }
3888: break;
3889: case 4:
3890: ivi = ivect->index.i;
3891: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3892: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3893: prev = pos;
3894: DMA(c1,c,c2,up,lo);
3895: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3896: } else svect[pos] = lo;
3897: }
3898: break;
1.65 noro 3899: }
3900: }
3901: }
1.66 noro 3902: for ( i = 0; i < col; i++ )
3903: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.65 noro 3904: }
3905:
3906: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3907: {
3908: int j,k,len;
3909: UINT *p;
3910: UINT c;
3911: NDV r;
3912: NMV mr0,mr;
3913:
3914: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3915: if ( !len ) return 0;
3916: else {
3917: mr0 = (NMV)MALLOC_ATOMIC(nmv_adv*len);
3918: mr = mr0;
3919: p = s0vect;
3920: for ( j = k = 0; j < col; j++, p += nd_wpd )
3921: if ( !rhead[j] ) {
3922: if ( c = vect[k++] ) {
3923: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3924: }
3925: }
3926: MKNDV(nd_nvar,mr0,len,r);
3927: return r;
3928: }
3929: }
3930:
1.68 ! noro 3931: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket,NODE *s)
1.65 noro 3932: {
3933: ND_pairs t;
3934: NODE sp0,sp;
3935: int stat;
3936: ND spol;
3937:
3938: sp0 = 0;
3939: for ( t = l; t; t = NEXT(t) ) {
3940: stat = nd_sp(m,0,t,&spol);
3941: if ( !stat ) return 0;
3942: if ( spol ) {
3943: NEXTNODE(sp0,sp); BDY(sp) = (pointer)nd_dup(spol);
3944: add_pbucket_symbolic(bucket,spol);
3945: }
3946: }
1.68 ! noro 3947: *s = sp0;
! 3948: return 1;
1.65 noro 3949: }
3950:
3951: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
3952: {
3953: NODE rp0,rp;
3954: NM mul,head,s0,s;
3955: int index,col,i;
3956: RHist h;
3957: UINT *s0v,*p;
3958: NM_ind_pair pair;
3959: ND red;
3960:
3961: s0 = 0; rp0 = 0; col = 0;
3962: while ( 1 ) {
3963: head = remove_head_pbucket_symbolic(bucket);
3964: if ( !head ) break;
3965: if ( !s0 ) s0 = head;
3966: else NEXT(s) = head;
3967: s = head;
3968: index = ndl_find_reducer(DL(head));
3969: if ( index >= 0 ) {
3970: h = nd_psh[index];
3971: NEWNM(mul);
3972: ndl_sub(DL(head),DL(h),DL(mul));
3973: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
3974: MKNM_ind_pair(pair,mul,index);
3975: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
3976: add_pbucket_symbolic(bucket,nd_remove_head(red));
3977: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
3978: }
3979: col++;
3980: }
3981: NEXT(rp) = 0; NEXT(s) = 0;
3982: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
3983: for ( i = 0, p = s0v, s = s0; i < col;
3984: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
3985: *s0vect = s0v;
3986: *r = rp0;
3987: return col;
3988: }
3989:
1.63 noro 3990: NODE nd_f4(int m)
3991: {
3992: int i,nh,stat,index;
3993: NODE r,g;
3994: ND_pairs d,l,t;
3995: ND spol,red;
1.65 noro 3996: NDV nf,redv;
1.63 noro 3997: NM s0,s;
3998: NODE sp0,sp,rp0,rp;
1.65 noro 3999: int nsp,nred,col,rank,len,k,j,a;
4000: UINT c;
1.63 noro 4001: UINT **spmat;
1.65 noro 4002: UINT *s0vect,*svect,*p,*v;
1.63 noro 4003: int *colstat;
1.67 noro 4004: IndArray *imat;
1.65 noro 4005: int *rhead;
4006: int spcol,sprow;
4007: int sugar;
1.63 noro 4008: PGeoBucket bucket;
1.66 noro 4009: struct oEGT eg0,eg1,eg_f4;
1.63 noro 4010:
4011: if ( !m )
4012: error("nd_f4 : not implemented");
4013:
4014: g = 0; d = 0;
4015: for ( i = 0; i < nd_psn; i++ ) {
4016: d = update_pairs(d,g,i);
4017: g = update_base(g,i);
4018: }
4019: while ( d ) {
1.66 noro 4020: get_eg(&eg0);
1.63 noro 4021: l = nd_minsugarp(d,&d);
4022: sugar = SG(l);
4023: bucket = create_pbucket();
1.68 ! noro 4024: stat = nd_sp_f4(m,l,bucket,&sp0);
! 4025: if ( !stat ) {
! 4026: for ( t = l; NEXT(t); t = NEXT(t) );
! 4027: NEXT(t) = d; d = l;
! 4028: d = nd_reconstruct(m,0,d);
! 4029: continue;
! 4030: }
! 4031: if ( !sp0 ) continue;
! 4032: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
! 4033: if ( !col ) {
1.65 noro 4034: for ( t = l; NEXT(t); t = NEXT(t) );
4035: NEXT(t) = d; d = l;
4036: d = nd_reconstruct(m,0,d);
4037: continue;
4038: }
4039:
4040: nsp = length(sp0); nred = length(rp0); spcol = col-nred;
1.67 noro 4041: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
1.65 noro 4042: rhead = (int *)MALLOC_ATOMIC(col*sizeof(int));
4043: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4044:
4045: /* construction of index arrays */
4046: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
1.67 noro 4047: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,(NM_ind_pair)BDY(rp));
4048: rhead[imat[i]->head] = 1;
1.65 noro 4049: }
4050:
4051: /* elimination (1st step) */
1.63 noro 4052: spmat = (UINT **)MALLOC(nsp*sizeof(UINT *));
1.65 noro 4053: svect = (UINT *)MALLOC_ATOMIC(col*sizeof(UINT));
4054: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4055: nd_to_vect(m,s0vect,col,BDY(sp),svect);
1.66 noro 4056: ndv_reduce_vect(m,svect,col,imat,rp0);
1.65 noro 4057: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4058: if ( i < col ) {
4059: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4060: for ( j = k = 0; j < col; j++ )
4061: if ( !rhead[j] ) v[k++] = svect[j];
4062: sprow++;
4063: }
4064: }
4065: /* free index arrays */
1.67 noro 4066: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
1.65 noro 4067:
4068: /* elimination (2nd step) */
4069: colstat = (int *)ALLOCA(spcol*sizeof(int));
4070: rank = generic_gauss_elim_mod(spmat,sprow,spcol,m,colstat);
4071:
1.66 noro 4072: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4073: fprintf(asir_out,"sugar=%d,nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4074: sugar,nsp,nred,sprow,spcol,rank);
4075: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.65 noro 4076:
4077: /* adding new bases */
1.66 noro 4078: for ( i = 0; i < rank; i++ ) {
1.65 noro 4079: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4080: SG(nf) = sugar;
4081: ndv_removecont(m,nf);
4082: nh = ndv_newps(nf,0);
4083: d = update_pairs(d,g,nh);
4084: g = update_base(g,nh);
4085: GC_free(spmat[i]);
1.63 noro 4086: }
1.66 noro 4087: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.63 noro 4088: }
4089: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
4090: return g;
1.59 noro 4091: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>