Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.81
1.81 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.80 2003/10/10 10:07:18 noro Exp $ */
1.2 noro 2:
1.1 noro 3: #include "ca.h"
1.74 noro 4: #include "parse.h"
5: #include "ox.h"
1.1 noro 6: #include "inline.h"
1.64 noro 7: #include <time.h>
1.1 noro 8:
9: #if defined(__GNUC__)
10: #define INLINE inline
11: #elif defined(VISUAL)
12: #define INLINE __inline
13: #else
14: #define INLINE
15: #endif
16:
1.61 noro 17: typedef unsigned int UINT;
18:
1.47 noro 19: #define USE_GEOBUCKET 1
1.65 noro 20: #define USE_UNROLL 1
1.28 noro 21:
1.1 noro 22: #define REDTAB_LEN 32003
23:
1.40 noro 24: /* GeoBucket for polynomial addition */
25:
1.1 noro 26: typedef struct oPGeoBucket {
27: int m;
28: struct oND *body[32];
29: } *PGeoBucket;
30:
1.40 noro 31: /* distributed polynomial; linked list rep. */
1.1 noro 32: typedef struct oND {
33: struct oNM *body;
34: int nv;
1.31 noro 35: int len;
1.1 noro 36: int sugar;
37: } *ND;
38:
1.40 noro 39: /* distributed polynomial; array rep. */
1.3 noro 40: typedef struct oNDV {
41: struct oNMV *body;
42: int nv;
1.31 noro 43: int len;
1.3 noro 44: int sugar;
45: } *NDV;
46:
1.69 noro 47: typedef union oNDC {
48: int m;
49: Q z;
50: P p;
51: } *NDC;
52:
1.40 noro 53: /* monomial; linked list rep. */
1.1 noro 54: typedef struct oNM {
55: struct oNM *next;
1.69 noro 56: union oNDC c;
1.61 noro 57: UINT dl[1];
1.1 noro 58: } *NM;
59:
1.40 noro 60: /* monomial; array rep. */
1.3 noro 61: typedef struct oNMV {
1.69 noro 62: union oNDC 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;
1.76 noro 103: int index,sugar;
1.63 noro 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:
1.74 noro 119: static int ndv_alloc;
1.69 noro 120: static int nd_f4_nsp=0x7fffffff;
1.42 noro 121: static double nd_scale=2;
1.61 noro 122: static UINT **nd_bound;
1.42 noro 123: static struct order_spec *nd_ord;
124: static EPOS nd_epos;
1.43 noro 125: static BlockMask nd_blockmask;
1.42 noro 126: static int nd_nvar;
127: static int nd_isrlex;
128: static int nd_epw,nd_bpe,nd_wpd,nd_exporigin;
1.61 noro 129: static UINT nd_mask[32];
130: static UINT nd_mask0,nd_mask1;
1.42 noro 131:
132: static NM _nm_free_list;
133: static ND _nd_free_list;
134: static ND_pairs _ndp_free_list;
1.20 noro 135:
136: static NDV *nd_ps;
1.53 noro 137: static NDV *nd_ps_trace;
1.42 noro 138: static RHist *nd_psh;
139: static int nd_psn,nd_pslen;
1.20 noro 140:
1.42 noro 141: static RHist *nd_red;
1.1 noro 142:
1.42 noro 143: static int nd_found,nd_create,nd_notfirst;
144: static int nm_adv;
145: static int nmv_adv;
146: static int nd_dcomp;
1.77 noro 147: static int nd_demand;
1.1 noro 148:
1.74 noro 149: extern struct order_spec dp_current_spec;
1.77 noro 150: extern char *Demand;
1.61 noro 151: extern VL CO;
1.77 noro 152: extern int Top,Reverse,DP_Print,dp_nelim,do_weyl,NoSugar;
1.58 noro 153: extern int *current_weyl_weight_vector;
1.1 noro 154:
1.40 noro 155: /* fundamental macros */
1.34 noro 156: #define TD(d) (d[0])
1.1 noro 157: #define HDL(d) ((d)->body->dl)
1.34 noro 158: #define HTD(d) (TD(HDL(d)))
1.14 noro 159: #define HCM(d) ((d)->body->c.m)
1.16 noro 160: #define HCQ(d) ((d)->body->c.z)
1.61 noro 161: #define HCP(d) ((d)->body->c.p)
1.14 noro 162: #define CM(a) ((a)->c.m)
1.16 noro 163: #define CQ(a) ((a)->c.z)
1.61 noro 164: #define CP(a) ((a)->c.p)
1.14 noro 165: #define DL(a) ((a)->dl)
166: #define SG(a) ((a)->sugar)
167: #define LEN(a) ((a)->len)
1.33 noro 168: #define LCM(a) ((a)->lcm)
1.42 noro 169: #define GET_EXP(d,a) (((d)[nd_epos[a].i]>>nd_epos[a].s)&nd_mask0)
1.60 noro 170: #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 171: #define PUT_EXP(r,a,e) ((r)[nd_epos[a].i] |= ((e)<<nd_epos[a].s))
1.45 noro 172: #define XOR_EXP(r,a,e) ((r)[nd_epos[a].i] ^= ((e)<<nd_epos[a].s))
1.1 noro 173:
1.61 noro 174: #define GET_EXP_OLD(d,a) (((d)[oepos[a].i]>>oepos[a].s)&omask0)
175: #define PUT_EXP_OLD(r,a,e) ((r)[oepos[a].i] |= ((e)<<oepos[a].s))
176:
1.40 noro 177: /* macros for term comparison */
1.34 noro 178: #define TD_DL_COMPARE(d1,d2)\
1.41 noro 179: (TD(d1)>TD(d2)?1:(TD(d1)<TD(d2)?-1:ndl_lex_compare(d1,d2)))
1.43 noro 180: #if 0
1.34 noro 181: #define DL_COMPARE(d1,d2)\
1.43 noro 182: (nd_dcomp>0?TD_DL_COMPARE(d1,d2)\
183: :(nd_dcomp==0?ndl_lex_compare(d1,d2)\
184: :(nd_blockmask?ndl_block_compare(d1,d2)\
1.45 noro 185: :(*ndl_compare_function)(d1,d2))))
1.43 noro 186: #else
187: #define DL_COMPARE(d1,d2)\
1.45 noro 188: (nd_dcomp>0?TD_DL_COMPARE(d1,d2):(*ndl_compare_function)(d1,d2))
1.43 noro 189: #endif
1.34 noro 190:
1.40 noro 191: /* allocators */
1.15 noro 192: #define NEWRHist(r) \
1.61 noro 193: ((r)=(RHist)MALLOC(sizeof(struct oRHist)+(nd_wpd-1)*sizeof(UINT)))
1.34 noro 194: #define NEWND_pairs(m) \
195: if(!_ndp_free_list)_NDP_alloc();\
196: (m)=_ndp_free_list; _ndp_free_list = NEXT(_ndp_free_list)
197: #define NEWNM(m)\
198: if(!_nm_free_list)_NM_alloc();\
199: (m)=_nm_free_list; _nm_free_list = NEXT(_nm_free_list)
200: #define MKND(n,m,len,d)\
201: if(!_nd_free_list)_ND_alloc();\
202: (d)=_nd_free_list; _nd_free_list = (ND)BDY(_nd_free_list);\
203: NV(d)=(n); LEN(d)=(len); BDY(d)=(m)
1.40 noro 204: #define NEWNDV(d) ((d)=(NDV)MALLOC(sizeof(struct oNDV)))
205: #define MKNDV(n,m,l,d) NEWNDV(d); NV(d)=(n); BDY(d)=(m); LEN(d) = l;
1.63 noro 206: #define NEWNM_ind_pair(p)\
207: ((p)=(NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair)))
1.1 noro 208:
1.40 noro 209: /* allocate and link a new object */
1.13 noro 210: #define NEXTRHist(r,c) \
211: if(!(r)){NEWRHist(r);(c)=(r);}else{NEWRHist(NEXT(c));(c)=NEXT(c);}
1.1 noro 212: #define NEXTNM(r,c) \
213: if(!(r)){NEWNM(r);(c)=(r);}else{NEWNM(NEXT(c));(c)=NEXT(c);}
214: #define NEXTNM2(r,c,s) \
215: if(!(r)){(c)=(r)=(s);}else{NEXT(c)=(s);(c)=(s);}
1.40 noro 216: #define NEXTND_pairs(r,c) \
217: if(!(r)){NEWND_pairs(r);(c)=(r);}else{NEWND_pairs(NEXT(c));(c)=NEXT(c);}
1.76 noro 218: #define MKNM_ind_pair(p,m,i,s) (NEWNM_ind_pair(p),(p)->mul=(m),(p)->index=(i),(p)->sugar = (s))
1.34 noro 219:
1.40 noro 220: /* deallocators */
1.1 noro 221: #define FREENM(m) NEXT(m)=_nm_free_list; _nm_free_list=(m)
222: #define FREENDP(m) NEXT(m)=_ndp_free_list; _ndp_free_list=(m)
223: #define FREEND(m) BDY(m)=(NM)_nd_free_list; _nd_free_list=(m)
224:
1.40 noro 225: /* macro for increasing pointer to NMV */
226: #define NMV_ADV(m) (m = (NMV)(((char *)m)+nmv_adv))
1.61 noro 227: #define NMV_OADV(m) (m = (NMV)(((char *)m)+oadv))
228: #define NDV_NADV(m) (m = (NMV)(((char *)m)+newadv))
1.56 noro 229: #define NMV_PREV(m) (m = (NMV)(((char *)m)-nmv_adv))
1.61 noro 230: #define NMV_OPREV(m) (m = (NMV)(((char *)m)-oadv))
231:
1.40 noro 232: /* external functions */
1.71 noro 233: #if 1
1.40 noro 234: void GC_gcollect();
1.71 noro 235: #endif
1.40 noro 236: NODE append_one(NODE,int);
1.1 noro 237:
1.40 noro 238: /* manipulation of coefficients */
1.20 noro 239: void nd_removecont(int mod,ND p);
1.21 noro 240: void nd_removecont2(ND p1,ND p2);
1.40 noro 241: void removecont_array(Q *c,int n);
242:
243: /* GeoBucket functions */
1.25 noro 244: ND normalize_pbucket(int mod,PGeoBucket g);
245: int head_pbucket(int mod,PGeoBucket g);
1.26 noro 246: int head_pbucket_q(PGeoBucket g);
1.63 noro 247: void add_pbucket_symbolic(PGeoBucket g,ND d);
1.31 noro 248: void add_pbucket(int mod,PGeoBucket g,ND d);
1.25 noro 249: void free_pbucket(PGeoBucket b);
1.26 noro 250: void mulq_pbucket(PGeoBucket g,Q c);
1.63 noro 251: NM remove_head_pbucket_symbolic(PGeoBucket g);
1.25 noro 252: PGeoBucket create_pbucket();
1.20 noro 253:
1.40 noro 254: /* manipulation of pairs and bases */
1.39 noro 255: int nd_newps(int mod,ND a,ND aq);
1.40 noro 256: ND_pairs nd_newpairs( NODE g, int t );
1.1 noro 257: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest );
1.63 noro 258: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest );
1.1 noro 259: NODE update_base(NODE nd,int ndp);
1.40 noro 260: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t);
261: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest );
262: ND_pairs crit_B( ND_pairs d, int s );
263: ND_pairs crit_M( ND_pairs d1 );
264: ND_pairs crit_F( ND_pairs d1 );
1.1 noro 265: int crit_2( int dp1, int dp2 );
1.77 noro 266: int ndv_newps(int m,NDV a,NDV aq);
1.40 noro 267:
268: /* top level functions */
1.63 noro 269: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp);
1.52 noro 270: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp);
1.63 noro 271: NODE nd_f4(int m);
1.77 noro 272: NODE nd_gb(int m,int ishomo,int checkonly);
1.23 noro 273: NODE nd_gb_trace(int m);
1.40 noro 274:
275: /* ndl functions */
1.61 noro 276: int ndl_weight(UINT *d);
277: void ndl_weight_mask(UINT *d);
278: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight);
279: void ndl_dehomogenize(UINT *p);
280: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos);
281: INLINE int ndl_reducible(UINT *d1,UINT *d2);
282: INLINE int ndl_lex_compare(UINT *d1,UINT *d2);
283: INLINE int ndl_block_compare(UINT *d1,UINT *d2);
284: INLINE int ndl_equal(UINT *d1,UINT *d2);
285: INLINE void ndl_copy(UINT *d1,UINT *d2);
286: INLINE void ndl_zero(UINT *d);
287: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d);
288: INLINE void ndl_addto(UINT *d1,UINT *d2);
289: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d);
290: INLINE int ndl_hash_value(UINT *d);
1.45 noro 291:
292: /* normal forms */
1.63 noro 293: INLINE int ndl_find_reducer(UINT *g);
1.53 noro 294: int nd_sp(int mod,int trace,ND_pairs p,ND *nf);
1.69 noro 295: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *nf);
1.53 noro 296: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *nf);
1.40 noro 297:
298: /* finalizers */
1.61 noro 299: NODE ndv_reducebase(NODE x);
300: NODE ndv_reduceall(int m,NODE f);
1.40 noro 301:
302: /* allocators */
303: void nd_free_private_storage();
304: void _NM_alloc();
305: void _ND_alloc();
1.1 noro 306: void nd_free(ND p);
1.40 noro 307: void nd_free_redlist();
308:
309: /* printing */
1.61 noro 310: void ndl_print(UINT *dl);
1.1 noro 311: void nd_print(ND p);
1.16 noro 312: void nd_print_q(ND p);
1.1 noro 313: void ndp_print(ND_pairs d);
1.40 noro 314:
315:
316: /* setup, reconstruct */
317: void nd_init_ord(struct order_spec *spec);
318: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs ndp);
1.61 noro 319: void ndv_setup(int mod,int trace,NODE f);
320: void nd_setup_parameters(int nvar,int max);
1.43 noro 321: BlockMask nd_create_blockmask(struct order_spec *ord);
1.57 noro 322: EPOS nd_create_epos(struct order_spec *ord);
1.48 noro 323: int nd_get_exporigin(struct order_spec *ord);
1.61 noro 324: void ndv_mod(int mod,NDV p);
325: NDV ndv_dup(int mod,NDV p);
1.63 noro 326: ND nd_dup(ND p);
1.40 noro 327:
328: /* ND functions */
1.61 noro 329: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand);
1.40 noro 330: void nd_mul_c(int mod,ND p,int mul);
331: void nd_mul_c_q(ND p,Q mul);
1.61 noro 332: void nd_mul_c_p(VL vl,ND p,P mul);
1.40 noro 333: ND nd_remove_head(ND p);
1.69 noro 334: ND nd_separate_head(ND p,ND *head);
1.1 noro 335: int nd_length(ND p);
1.61 noro 336: void nd_append_red(UINT *d,int i);
337: UINT *ndv_compute_bound(NDV p);
1.6 noro 338: ND nd_copy(ND p);
1.63 noro 339: ND nd_merge(ND p1,ND p2);
1.40 noro 340: ND nd_add(int mod,ND p1,ND p2);
341: ND nd_add_q(ND p1,ND p2);
1.71 noro 342: ND nd_add_sf(ND p1,ND p2);
1.41 noro 343: INLINE int nd_length(ND p);
1.74 noro 344: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
345: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
1.4 noro 346:
1.40 noro 347: /* NDV functions */
1.55 noro 348: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p);
349: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen);
1.19 noro 350: void ndv_mul_c(int mod,NDV p,int mul);
1.40 noro 351: void ndv_mul_c_q(NDV p,Q mul);
1.63 noro 352: ND ndv_mul_nm_symbolic(NM m0,NDV p);
1.61 noro 353: ND ndv_mul_nm(int mod,NM m0,NDV p);
1.43 noro 354: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos);
1.61 noro 355: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos);
356: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS eops);
1.45 noro 357: void ndv_dehomogenize(NDV p,struct order_spec *spec);
1.40 noro 358: void ndv_removecont(int mod,NDV p);
359: void ndv_print(NDV p);
360: void ndv_print_q(NDV p);
361: void ndv_free(NDV p);
1.77 noro 362: void ndv_save(NDV p,int index);
363: NDV ndv_load(int index);
1.40 noro 364:
365: /* converters */
1.61 noro 366: ND ptond(VL vl,VL dvl,P p);
367: NDV ptondv(VL vl,VL dvl,P p);
368: P ndvtop(int mod,VL vl,VL dvl,NDV p);
1.16 noro 369: NDV ndtondv(int mod,ND p);
1.23 noro 370: ND ndvtond(int mod,NDV p);
1.63 noro 371: int nm_ind_pair_to_vect(int m,UINT *s0,int n,NM_ind_pair pair,UINT *r);
1.67 noro 372: IndArray nm_ind_pair_to_vect_compress(int m,UINT *s0,int n,NM_ind_pair pair);
1.63 noro 373: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r);
1.1 noro 374:
1.76 noro 375: /* elimination */
376: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat);
377: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat);
378:
1.1 noro 379: void nd_free_private_storage()
380: {
381: _nm_free_list = 0;
1.5 noro 382: _ndp_free_list = 0;
1.71 noro 383: #if 0
1.1 noro 384: GC_gcollect();
1.71 noro 385: #endif
1.1 noro 386: }
387:
388: void _NM_alloc()
389: {
390: NM p;
391: int i;
392:
1.11 noro 393: for ( i = 0; i < 1024; i++ ) {
1.61 noro 394: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 395: p->next = _nm_free_list; _nm_free_list = p;
396: }
397: }
398:
399: void _ND_alloc()
400: {
401: ND p;
402: int i;
403:
404: for ( i = 0; i < 1024; i++ ) {
405: p = (ND)GC_malloc(sizeof(struct oND));
406: p->body = (NM)_nd_free_list; _nd_free_list = p;
407: }
408: }
409:
410: void _NDP_alloc()
411: {
412: ND_pairs p;
413: int i;
414:
1.11 noro 415: for ( i = 0; i < 1024; i++ ) {
1.1 noro 416: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 417: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 418: p->next = _ndp_free_list; _ndp_free_list = p;
419: }
420: }
421:
1.30 noro 422: INLINE int nd_length(ND p)
1.1 noro 423: {
424: NM m;
425: int i;
426:
427: if ( !p )
428: return 0;
429: else {
430: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
431: return i;
432: }
433: }
434:
1.61 noro 435: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 436: {
1.61 noro 437: UINT u1,u2;
1.1 noro 438: int i,j;
439:
1.34 noro 440: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 441: #if USE_UNROLL
1.1 noro 442: switch ( nd_bpe ) {
1.62 noro 443: case 3:
444: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
445: u1 = d1[i]; u2 = d2[i];
446: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
447: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
448: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
449: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
450: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
451: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
452: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
453: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
454: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
455: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
456: }
457: return 1;
458: break;
1.1 noro 459: case 4:
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&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 463: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
464: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
465: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
466: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
467: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
468: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
469: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 470: }
471: return 1;
472: break;
473: case 6:
1.41 noro 474: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 475: u1 = d1[i]; u2 = d2[i];
476: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 477: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
478: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
479: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
480: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 481: }
482: return 1;
483: break;
484: case 8:
1.41 noro 485: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 486: u1 = d1[i]; u2 = d2[i];
487: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 488: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
489: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
490: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 491: }
492: return 1;
493: break;
494: case 16:
1.41 noro 495: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 496: u1 = d1[i]; u2 = d2[i];
497: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 498: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 499: }
500: return 1;
501: break;
502: case 32:
1.41 noro 503: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 504: if ( d1[i] < d2[i] ) return 0;
505: return 1;
506: break;
507: default:
1.41 noro 508: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 509: u1 = d1[i]; u2 = d2[i];
510: for ( j = 0; j < nd_epw; j++ )
511: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
512: }
513: return 1;
514: }
1.65 noro 515: #else
516: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
517: u1 = d1[i]; u2 = d2[i];
518: for ( j = 0; j < nd_epw; j++ )
519: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
520: }
521: return 1;
522: #endif
1.1 noro 523: }
524:
1.61 noro 525: /*
526: * If the current order is a block order,
527: * then the last block is length 1 and contains
528: * the homo variable. Otherwise, the original
529: * order is either 0 or 2.
530: */
531:
532: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 533: {
1.61 noro 534: int w,i,e,n,omask0;
535:
536: omask0 = (1<<obpe)-1;
537: n = nd_nvar-1;
538: ndl_zero(r);
539: for ( i = 0; i < n; i++ ) {
540: e = GET_EXP_OLD(d,i);
541: PUT_EXP(r,i,e);
542: }
543: w = TD(d);
544: PUT_EXP(r,nd_nvar-1,weight-w);
545: TD(r) = weight;
546: if ( nd_blockmask ) ndl_weight_mask(r);
547: }
548:
549: void ndl_dehomogenize(UINT *d)
550: {
551: UINT mask;
552: UINT h;
1.31 noro 553: int i,bits;
1.23 noro 554:
1.44 noro 555: if ( nd_blockmask ) {
556: h = GET_EXP(d,nd_nvar-1);
1.45 noro 557: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 558: TD(d) -= h;
559: d[nd_exporigin-1] -= h;
560: } else {
561: if ( nd_isrlex ) {
562: if ( nd_bpe == 32 ) {
563: h = d[nd_exporigin];
564: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
565: d[i-1] = d[i];
566: d[i-1] = 0;
567: TD(d) -= h;
568: } else {
569: bits = nd_epw*nd_bpe;
570: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
571: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
572: for ( i = nd_exporigin; i < nd_wpd; i++ )
573: d[i] = ((d[i]<<nd_bpe)&mask)
574: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
575: TD(d) -= h;
576: }
1.45 noro 577: } else {
578: h = GET_EXP(d,nd_nvar-1);
579: XOR_EXP(d,nd_nvar-1,h);
580: TD(d) -= h;
581: }
1.44 noro 582: }
1.23 noro 583: }
584:
1.61 noro 585: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 586: {
1.61 noro 587: UINT t1,t2,u,u1,u2;
1.43 noro 588: int i,j,l;
1.1 noro 589:
1.65 noro 590: #if USE_UNROLL
1.1 noro 591: switch ( nd_bpe ) {
1.62 noro 592: case 3:
593: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
594: u1 = d1[i]; u2 = d2[i];
595: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
596: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
597: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
598: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
599: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
600: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
601: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
602: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
603: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
604: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
605: d[i] = u;
606: }
607: break;
1.1 noro 608: case 4:
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&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 612: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
613: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
614: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
615: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
616: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
617: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
618: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 619: d[i] = u;
620: }
621: break;
622: case 6:
1.41 noro 623: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 624: u1 = d1[i]; u2 = d2[i];
625: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 626: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
627: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
628: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
629: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 630: d[i] = u;
631: }
632: break;
633: case 8:
1.41 noro 634: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 635: u1 = d1[i]; u2 = d2[i];
636: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 637: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
638: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
639: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 640: d[i] = u;
641: }
642: break;
643: case 16:
1.41 noro 644: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 645: u1 = d1[i]; u2 = d2[i];
646: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 647: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 648: d[i] = u;
649: }
650: break;
651: case 32:
1.41 noro 652: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 653: u1 = d1[i]; u2 = d2[i];
654: d[i] = u1>u2?u1:u2;
655: }
656: break;
657: default:
1.41 noro 658: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 659: u1 = d1[i]; u2 = d2[i];
660: for ( j = 0, u = 0; j < nd_epw; j++ ) {
661: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
662: }
663: d[i] = u;
664: }
665: break;
666: }
1.65 noro 667: #else
668: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
669: u1 = d1[i]; u2 = d2[i];
670: for ( j = 0, u = 0; j < nd_epw; j++ ) {
671: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
672: }
673: d[i] = u;
674: }
675: #endif
1.39 noro 676: TD(d) = ndl_weight(d);
1.61 noro 677: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 678: }
679:
1.61 noro 680: int ndl_weight(UINT *d)
1.1 noro 681: {
1.61 noro 682: UINT t,u;
1.1 noro 683: int i,j;
684:
1.60 noro 685: if ( current_dl_weight_vector )
686: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
687: u = GET_EXP(d,i);
688: t += MUL_WEIGHT(u,i);
689: }
690: else
691: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
692: u = d[i];
693: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
694: t += (u&nd_mask0);
695: }
1.1 noro 696: return t;
697: }
698:
1.61 noro 699: void ndl_weight_mask(UINT *d)
1.43 noro 700: {
1.61 noro 701: UINT t,u;
702: UINT *mask;
703: int i,j,k,l;
1.43 noro 704:
1.61 noro 705: l = nd_blockmask->n;
706: for ( k = 0; k < l; k++ ) {
707: mask = nd_blockmask->mask[k];
708: if ( current_dl_weight_vector )
709: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
710: u = GET_EXP_MASK(d,i,mask);
711: t += MUL_WEIGHT(u,i);
712: }
713: else
714: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
715: u = d[i]&mask[i];
716: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
717: t += (u&nd_mask0);
718: }
719: d[k+1] = t;
720: }
1.43 noro 721: }
722:
1.61 noro 723: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 724: {
725: int i;
726:
1.41 noro 727: d1 += nd_exporigin;
728: d2 += nd_exporigin;
729: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 730: if ( *d1 > *d2 )
1.32 noro 731: return nd_isrlex ? -1 : 1;
1.1 noro 732: else if ( *d1 < *d2 )
1.32 noro 733: return nd_isrlex ? 1 : -1;
1.1 noro 734: return 0;
735: }
736:
1.61 noro 737: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 738: {
739: int i,l,j,ord_o,ord_l;
740: struct order_pair *op;
1.61 noro 741: UINT t1,t2,m;
742: UINT *mask;
1.43 noro 743:
744: l = nd_blockmask->n;
745: op = nd_blockmask->order_pair;
746: for ( j = 0; j < l; j++ ) {
747: mask = nd_blockmask->mask[j];
748: ord_o = op[j].order;
749: if ( ord_o < 2 )
1.44 noro 750: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
751: else if ( t1 < t2 ) return -1;
1.43 noro 752: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 753: m = mask[i];
754: t1 = d1[i]&m;
755: t2 = d2[i]&m;
1.43 noro 756: if ( t1 > t2 )
757: return !ord_o ? -1 : 1;
758: else if ( t1 < t2 )
759: return !ord_o ? 1 : -1;
760: }
761: }
762: return 0;
763: }
764:
1.58 noro 765: /* TDH -> WW -> TD-> RL */
766:
1.61 noro 767: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 768: {
769: int i,m,e1,e2;
770:
771: if ( TD(d1) > TD(d2) ) return 1;
772: else if ( TD(d1) < TD(d2) ) return -1;
773: m = nd_nvar>>1;
774: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
775: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
776: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
777: }
778: if ( e1 > e2 ) return 1;
779: else if ( e1 < e2 ) return -1;
780: return ndl_lex_compare(d1,d2);
781: }
782:
1.61 noro 783: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 784: {
785: int i;
786:
1.81 ! noro 787: switch ( nd_wpd ) {
! 788: case 2:
! 789: if ( TD(d2) != TD(d1) ) return 0;
! 790: if ( d2[1] != d1[1] ) return 0;
! 791: return 1;
! 792: break;
! 793: case 3:
! 794: if ( TD(d2) != TD(d1) ) return 0;
! 795: if ( d2[1] != d1[1] ) return 0;
! 796: if ( d2[2] != d1[2] ) return 0;
! 797: return 1;
! 798: break;
! 799: default:
! 800: for ( i = 0; i < nd_wpd; i++ )
! 801: if ( *d1++ != *d2++ ) return 0;
! 802: return 1;
! 803: break;
! 804: }
1.1 noro 805: }
806:
1.61 noro 807: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 808: {
809: int i;
810:
811: switch ( nd_wpd ) {
1.41 noro 812: case 2:
1.34 noro 813: TD(d2) = TD(d1);
814: d2[1] = d1[1];
1.6 noro 815: break;
1.41 noro 816: case 3:
1.34 noro 817: TD(d2) = TD(d1);
1.6 noro 818: d2[1] = d1[1];
1.34 noro 819: d2[2] = d1[2];
1.6 noro 820: break;
821: default:
1.41 noro 822: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 823: d2[i] = d1[i];
824: break;
825: }
826: }
827:
1.61 noro 828: INLINE void ndl_zero(UINT *d)
829: {
830: int i;
831: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
832: }
833:
834: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 835: {
836: int i;
837:
1.43 noro 838: #if 1
1.6 noro 839: switch ( nd_wpd ) {
1.41 noro 840: case 2:
841: TD(d) = TD(d1)+TD(d2);
1.34 noro 842: d[1] = d1[1]+d2[1];
1.6 noro 843: break;
1.41 noro 844: case 3:
845: TD(d) = TD(d1)+TD(d2);
1.6 noro 846: d[1] = d1[1]+d2[1];
1.34 noro 847: d[2] = d1[2]+d2[2];
1.6 noro 848: break;
849: default:
1.43 noro 850: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 851: break;
852: }
1.43 noro 853: #else
854: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
855: #endif
1.6 noro 856: }
857:
1.55 noro 858: /* d1 += d2 */
1.61 noro 859: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 860: {
861: int i;
862:
863: #if 1
864: switch ( nd_wpd ) {
865: case 2:
866: TD(d1) += TD(d2);
867: d1[1] += d2[1];
868: break;
869: case 3:
870: TD(d1) += TD(d2);
871: d1[1] += d2[1];
872: d1[2] += d2[2];
873: break;
874: default:
875: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
876: break;
877: }
878: #else
879: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
880: #endif
881: }
882:
1.61 noro 883: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 884: {
885: int i;
886:
1.43 noro 887: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 888: }
889:
1.61 noro 890: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 891: {
1.61 noro 892: UINT t1,t2,u,u1,u2;
1.1 noro 893: int i,j;
894:
1.65 noro 895: #if USE_UNROLL
1.1 noro 896: switch ( nd_bpe ) {
1.62 noro 897: case 3:
898: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
899: u1 = d1[i]; u2 = d2[i];
900: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
901: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
902: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
903: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
904: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
905: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
906: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
907: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
908: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
909: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
910: }
911: return 1;
912: break;
1.1 noro 913: case 4:
1.41 noro 914: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 915: u1 = d1[i]; u2 = d2[i];
916: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 917: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
918: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
919: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
920: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
921: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
922: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
923: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 924: }
925: return 1;
926: break;
927: case 6:
1.41 noro 928: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 929: u1 = d1[i]; u2 = d2[i];
930: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 931: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
932: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
933: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
934: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 935: }
936: return 1;
937: break;
938: case 8:
1.41 noro 939: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 940: u1 = d1[i]; u2 = d2[i];
941: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 942: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
943: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
944: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 945: }
946: return 1;
947: break;
948: case 16:
1.41 noro 949: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 950: u1 = d1[i]; u2 = d2[i];
951: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 952: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 953: }
954: return 1;
955: break;
956: case 32:
1.41 noro 957: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 958: if ( d1[i] && d2[i] ) return 0;
959: return 1;
960: break;
961: default:
1.41 noro 962: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 963: u1 = d1[i]; u2 = d2[i];
964: for ( j = 0; j < nd_epw; j++ ) {
965: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
966: u1 >>= nd_bpe; u2 >>= nd_bpe;
967: }
968: }
969: return 1;
970: break;
971: }
1.65 noro 972: #else
973: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
974: u1 = d1[i]; u2 = d2[i];
975: for ( j = 0; j < nd_epw; j++ ) {
976: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
977: u1 >>= nd_bpe; u2 >>= nd_bpe;
978: }
979: }
980: return 1;
981: #endif
1.1 noro 982: }
983:
1.61 noro 984: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 985: {
1.61 noro 986: UINT u2;
987: UINT *d1;
1.5 noro 988: int i,j,ind,k;
1.1 noro 989:
1.5 noro 990: d1 = nd_bound[index];
991: ind = 0;
1.65 noro 992: #if USE_UNROLL
1.5 noro 993: switch ( nd_bpe ) {
1.62 noro 994: case 3:
995: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
996: u2 = d2[i];
997: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
998: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
999: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
1000: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
1001: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
1002: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
1003: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
1004: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
1005: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
1006: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
1007: }
1008: return 0;
1009: break;
1.5 noro 1010: case 4:
1.41 noro 1011: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1012: u2 = d2[i];
1013: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
1014: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
1015: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
1016: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
1017: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
1018: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
1019: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1020: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1021: }
1022: return 0;
1023: break;
1024: case 6:
1.41 noro 1025: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1026: u2 = d2[i];
1027: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1028: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1029: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1030: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1031: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1032: }
1033: return 0;
1034: break;
1035: case 8:
1.41 noro 1036: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1037: u2 = d2[i];
1038: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1039: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1040: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1041: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1042: }
1043: return 0;
1044: break;
1045: case 16:
1.41 noro 1046: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1047: u2 = d2[i];
1048: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1049: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1050: }
1051: return 0;
1052: break;
1053: case 32:
1.41 noro 1054: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1055: if ( d1[i]+d2[i]<d1[i] ) return 1;
1056: return 0;
1057: break;
1058: default:
1.41 noro 1059: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1060: u2 = d2[i];
1061: k = (nd_epw-1)*nd_bpe;
1062: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1063: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1064: }
1065: return 0;
1066: break;
1067: }
1.65 noro 1068: #else
1069: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1070: u2 = d2[i];
1071: k = (nd_epw-1)*nd_bpe;
1072: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1073: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1074: }
1075: return 0;
1076: #endif
1.1 noro 1077: }
1078:
1.61 noro 1079: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1080: {
1081: int i;
1082: int r;
1083:
1.34 noro 1084: r = 0;
1.41 noro 1085: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1086: r = ((r<<16)+d[i])%REDTAB_LEN;
1087: return r;
1088: }
1089:
1.63 noro 1090: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1091: {
1.13 noro 1092: RHist r;
1.6 noro 1093: int d,k,i;
1.1 noro 1094:
1.63 noro 1095: d = ndl_hash_value(dg);
1.13 noro 1096: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1097: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1098: if ( k > 0 ) nd_notfirst++;
1099: nd_found++;
1.13 noro 1100: return r->index;
1.1 noro 1101: }
1102: }
1.13 noro 1103: if ( Reverse )
1104: for ( i = nd_psn-1; i >= 0; i-- ) {
1105: r = nd_psh[i];
1.34 noro 1106: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1107: nd_create++;
1.34 noro 1108: nd_append_red(dg,i);
1.13 noro 1109: return i;
1110: }
1111: }
1112: else
1113: for ( i = 0; i < nd_psn; i++ ) {
1114: r = nd_psh[i];
1.34 noro 1115: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1116: nd_create++;
1.34 noro 1117: nd_append_red(dg,i);
1.13 noro 1118: return i;
1119: }
1.1 noro 1120: }
1.6 noro 1121: return -1;
1.1 noro 1122: }
1123:
1.63 noro 1124: ND nd_merge(ND p1,ND p2)
1125: {
1126: int n,c;
1127: int t,can,td1,td2;
1128: ND r;
1129: NM m1,m2,mr0,mr,s;
1130:
1131: if ( !p1 ) return p2;
1132: else if ( !p2 ) return p1;
1133: else {
1134: can = 0;
1135: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1136: c = DL_COMPARE(DL(m1),DL(m2));
1137: switch ( c ) {
1138: case 0:
1139: s = m1; m1 = NEXT(m1);
1140: can++; NEXTNM2(mr0,mr,s);
1141: s = m2; m2 = NEXT(m2); FREENM(s);
1142: break;
1143: case 1:
1144: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1145: break;
1146: case -1:
1147: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1148: break;
1149: }
1150: }
1151: if ( !mr0 )
1152: if ( m1 ) mr0 = m1;
1153: else if ( m2 ) mr0 = m2;
1154: else return 0;
1155: else if ( m1 ) NEXT(mr) = m1;
1156: else if ( m2 ) NEXT(mr) = m2;
1157: else NEXT(mr) = 0;
1158: BDY(p1) = mr0;
1159: SG(p1) = MAX(SG(p1),SG(p2));
1160: LEN(p1) = LEN(p1)+LEN(p2)-can;
1161: FREEND(p2);
1162: return p1;
1163: }
1164: }
1165:
1.31 noro 1166: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1167: {
1168: int n,c;
1.34 noro 1169: int t,can,td1,td2;
1.1 noro 1170: ND r;
1171: NM m1,m2,mr0,mr,s;
1172:
1.34 noro 1173: if ( !p1 ) return p2;
1174: else if ( !p2 ) return p1;
1.71 noro 1175: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 1176: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1177: else {
1.30 noro 1178: can = 0;
1.1 noro 1179: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1180: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1181: switch ( c ) {
1182: case 0:
1.19 noro 1183: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1184: if ( t < 0 ) t += mod;
1.1 noro 1185: s = m1; m1 = NEXT(m1);
1186: if ( t ) {
1.34 noro 1187: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1188: } else {
1.34 noro 1189: can += 2; FREENM(s);
1.1 noro 1190: }
1191: s = m2; m2 = NEXT(m2); FREENM(s);
1192: break;
1193: case 1:
1194: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1195: break;
1196: case -1:
1197: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1198: break;
1199: }
1200: }
1201: if ( !mr0 )
1.34 noro 1202: if ( m1 ) mr0 = m1;
1203: else if ( m2 ) mr0 = m2;
1204: else return 0;
1205: else if ( m1 ) NEXT(mr) = m1;
1206: else if ( m2 ) NEXT(mr) = m2;
1207: else NEXT(mr) = 0;
1.1 noro 1208: BDY(p1) = mr0;
1.14 noro 1209: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1210: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1211: FREEND(p2);
1212: return p1;
1213: }
1214: }
1215:
1.31 noro 1216: ND nd_add_q(ND p1,ND p2)
1.17 noro 1217: {
1.30 noro 1218: int n,c,can;
1.17 noro 1219: ND r;
1220: NM m1,m2,mr0,mr,s;
1221: Q t;
1222:
1.34 noro 1223: if ( !p1 ) return p2;
1224: else if ( !p2 ) return p1;
1.31 noro 1225: else {
1.30 noro 1226: can = 0;
1.17 noro 1227: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1228: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1229: switch ( c ) {
1230: case 0:
1231: addq(CQ(m1),CQ(m2),&t);
1232: s = m1; m1 = NEXT(m1);
1233: if ( t ) {
1.34 noro 1234: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1235: } else {
1.34 noro 1236: can += 2; FREENM(s);
1.17 noro 1237: }
1238: s = m2; m2 = NEXT(m2); FREENM(s);
1239: break;
1240: case 1:
1241: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1242: break;
1243: case -1:
1244: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1245: break;
1246: }
1247: }
1248: if ( !mr0 )
1.34 noro 1249: if ( m1 ) mr0 = m1;
1250: else if ( m2 ) mr0 = m2;
1251: else return 0;
1252: else if ( m1 ) NEXT(mr) = m1;
1253: else if ( m2 ) NEXT(mr) = m2;
1254: else NEXT(mr) = 0;
1.17 noro 1255: BDY(p1) = mr0;
1256: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1257: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1258: FREEND(p2);
1259: return p1;
1260: }
1261: }
1262:
1.71 noro 1263: ND nd_add_sf(ND p1,ND p2)
1264: {
1265: int n,c,can;
1266: ND r;
1267: NM m1,m2,mr0,mr,s;
1268: int t;
1269:
1270: if ( !p1 ) return p2;
1271: else if ( !p2 ) return p1;
1272: else {
1273: can = 0;
1274: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1275: c = DL_COMPARE(DL(m1),DL(m2));
1276: switch ( c ) {
1277: case 0:
1278: t = _addsf(CM(m1),CM(m2));
1279: s = m1; m1 = NEXT(m1);
1280: if ( t ) {
1281: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1282: } else {
1283: can += 2; FREENM(s);
1284: }
1285: s = m2; m2 = NEXT(m2); FREENM(s);
1286: break;
1287: case 1:
1288: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1289: break;
1290: case -1:
1291: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1292: break;
1293: }
1294: }
1295: if ( !mr0 )
1296: if ( m1 ) mr0 = m1;
1297: else if ( m2 ) mr0 = m2;
1298: else return 0;
1299: else if ( m1 ) NEXT(mr) = m1;
1300: else if ( m2 ) NEXT(mr) = m2;
1301: else NEXT(mr) = 0;
1302: BDY(p1) = mr0;
1303: SG(p1) = MAX(SG(p1),SG(p2));
1304: LEN(p1) = LEN(p1)+LEN(p2)-can;
1305: FREEND(p2);
1306: return p1;
1307: }
1308: }
1309:
1.1 noro 1310: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1311: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1312: {
1.11 noro 1313: ND d;
1.1 noro 1314: NM m,mrd,tail;
1.7 noro 1315: NM mul;
1.10 noro 1316: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1317: int c,c1,c2,dummy;
1.17 noro 1318: RHist h;
1.11 noro 1319: NDV p,red;
1.69 noro 1320: Q cg,cred,gcd,tq,qq;
1.21 noro 1321: double hmag;
1.1 noro 1322:
1.69 noro 1323: if ( dn ) {
1324: if ( mod )
1325: dn->m = 1;
1326: else
1327: dn->z = ONE;
1328: }
1.1 noro 1329: if ( !g ) {
1330: *rp = 0;
1331: return 1;
1332: }
1.34 noro 1333: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1334:
1.14 noro 1335: sugar0 = sugar = SG(g);
1.1 noro 1336: n = NV(g);
1.61 noro 1337: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1338: for ( d = 0; g; ) {
1.63 noro 1339: index = ndl_find_reducer(HDL(g));
1.6 noro 1340: if ( index >= 0 ) {
1.17 noro 1341: h = nd_psh[index];
1342: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1343: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1344: nd_free(g); nd_free(d);
1345: return 0;
1346: }
1.77 noro 1347: if ( nd_demand )
1348: p = ndv_load(index);
1349: else
1350: p = ps[index];
1.71 noro 1351: if ( mod == -1 )
1352: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1353: else if ( mod ) {
1.19 noro 1354: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1355: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1356: } else {
1.17 noro 1357: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1358: chsgnq(cg,&CQ(mul));
1.20 noro 1359: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1360: if ( dn ) {
1361: mulq(dn->z,cred,&tq); dn->z = tq;
1362: }
1.16 noro 1363: }
1.55 noro 1364: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1365: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1366: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1367: tq = HCQ(g);
1.21 noro 1368: nd_removecont2(d,g);
1.69 noro 1369: if ( dn ) {
1370: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1371: }
1.21 noro 1372: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1373: }
1.1 noro 1374: } else if ( !full ) {
1375: *rp = g;
1376: return 1;
1377: } else {
1378: m = BDY(g);
1379: if ( NEXT(m) ) {
1.34 noro 1380: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1381: } else {
1382: FREEND(g); g = 0;
1383: }
1384: if ( d ) {
1.34 noro 1385: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1386: } else {
1.34 noro 1387: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1388: }
1389: }
1390: }
1.34 noro 1391: if ( d ) SG(d) = sugar;
1.1 noro 1392: *rp = d;
1393: return 1;
1394: }
1.28 noro 1395:
1.53 noro 1396: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1397: {
1398: int hindex,index;
1399: NDV p;
1400: ND u,d,red;
1401: NODE l;
1.31 noro 1402: NM mul,m,mrd,tail;
1.25 noro 1403: int sugar,psugar,n,h_reducible;
1404: PGeoBucket bucket;
1405: int c,c1,c2;
1.26 noro 1406: Q cg,cred,gcd,zzz;
1.25 noro 1407: RHist h;
1.28 noro 1408: double hmag,gmag;
1.77 noro 1409: int count = 0;
1410: int hcount = 0;
1.25 noro 1411:
1412: if ( !g ) {
1413: *rp = 0;
1414: return 1;
1415: }
1416: sugar = SG(g);
1417: n = NV(g);
1.34 noro 1418: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1419: bucket = create_pbucket();
1.31 noro 1420: add_pbucket(mod,bucket,g);
1.25 noro 1421: d = 0;
1.61 noro 1422: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1423: while ( 1 ) {
1.26 noro 1424: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1425: if ( hindex < 0 ) {
1.77 noro 1426: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1427: if ( d ) SG(d) = sugar;
1.25 noro 1428: *rp = d;
1429: return 1;
1430: }
1431: g = bucket->body[hindex];
1.63 noro 1432: index = ndl_find_reducer(HDL(g));
1.25 noro 1433: if ( index >= 0 ) {
1.77 noro 1434: count++;
1435: if ( !d ) hcount++;
1.25 noro 1436: h = nd_psh[index];
1437: ndl_sub(HDL(g),DL(h),DL(mul));
1438: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1439: nd_free(d);
1.25 noro 1440: free_pbucket(bucket);
1441: *rp = 0;
1442: return 0;
1443: }
1.53 noro 1444: p = ps[index];
1.71 noro 1445: if ( mod == -1 )
1446: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1447: else if ( mod ) {
1.25 noro 1448: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1449: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1450: } else {
1451: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1452: chsgnq(cg,&CQ(mul));
1.26 noro 1453: nd_mul_c_q(d,cred);
1454: mulq_pbucket(bucket,cred);
1455: g = bucket->body[hindex];
1.28 noro 1456: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1457: }
1.55 noro 1458: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1459: bucket->body[hindex] = nd_remove_head(g);
1460: red = nd_remove_head(red);
1.31 noro 1461: add_pbucket(mod,bucket,red);
1.34 noro 1462: psugar = SG(p)+TD(DL(mul));
1463: sugar = MAX(sugar,psugar);
1.28 noro 1464: if ( !mod && hmag && (gmag > hmag) ) {
1465: g = normalize_pbucket(mod,bucket);
1466: if ( !g ) {
1.34 noro 1467: if ( d ) SG(d) = sugar;
1.28 noro 1468: *rp = d;
1469: return 1;
1470: }
1471: nd_removecont2(d,g);
1472: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1473: add_pbucket(mod,bucket,g);
1.28 noro 1474: }
1.25 noro 1475: } else if ( !full ) {
1476: g = normalize_pbucket(mod,bucket);
1.34 noro 1477: if ( g ) SG(g) = sugar;
1.25 noro 1478: *rp = g;
1479: return 1;
1480: } else {
1481: m = BDY(g);
1482: if ( NEXT(m) ) {
1.34 noro 1483: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1484: } else {
1485: FREEND(g); g = 0;
1486: }
1487: bucket->body[hindex] = g;
1488: NEXT(m) = 0;
1489: if ( d ) {
1.34 noro 1490: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1491: } else {
1.34 noro 1492: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1493: }
1494: }
1495: }
1496: }
1.27 noro 1497:
1.61 noro 1498: /* input : list of NDV, cand : list of NDV */
1.28 noro 1499:
1.61 noro 1500: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1501: {
1502: int n,i,stat;
1503: ND nf,d;
1.61 noro 1504: NDV r;
1.45 noro 1505: NODE t,s;
1.69 noro 1506: union oNDC dn;
1.45 noro 1507:
1.61 noro 1508: ndv_setup(0,0,cand);
1.31 noro 1509: n = length(cand);
1.28 noro 1510:
1511: /* membercheck : list is a subset of Id(cand) ? */
1512: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1513: again:
1.61 noro 1514: if ( nd_bpe > obpe )
1515: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1516: else
1517: r = (NDV)BDY(t);
1518: d = ndvtond(0,r);
1.69 noro 1519: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1520: if ( !stat ) {
1521: nd_reconstruct(0,0,0);
1522: goto again;
1523: } else if ( nf ) return 0;
1.71 noro 1524: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1525: }
1.71 noro 1526: if ( DP_Print ) { printf("\n"); }
1.28 noro 1527: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1528: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1529: /* XXX */
1.23 noro 1530: return 1;
1531: }
1.1 noro 1532:
1533: ND nd_remove_head(ND p)
1534: {
1535: NM m;
1536:
1537: m = BDY(p);
1538: if ( !NEXT(m) ) {
1.34 noro 1539: FREEND(p); p = 0;
1.31 noro 1540: } else {
1.34 noro 1541: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1542: }
1.1 noro 1543: FREENM(m);
1544: return p;
1545: }
1546:
1.69 noro 1547: ND nd_separate_head(ND p,ND *head)
1548: {
1549: NM m,m0;
1550: ND r;
1551:
1552: m = BDY(p);
1553: if ( !NEXT(m) ) {
1554: *head = p; p = 0;
1555: } else {
1556: m0 = m;
1557: BDY(p) = NEXT(m); LEN(p)--;
1558: NEXT(m0) = 0;
1559: MKND(NV(p),m0,1,r);
1560: *head = r;
1561: }
1562: return p;
1563: }
1564:
1.1 noro 1565: PGeoBucket create_pbucket()
1566: {
1567: PGeoBucket g;
1568:
1569: g = CALLOC(1,sizeof(struct oPGeoBucket));
1570: g->m = -1;
1571: return g;
1572: }
1573:
1.25 noro 1574: void free_pbucket(PGeoBucket b) {
1575: int i;
1576:
1.26 noro 1577: for ( i = 0; i <= b->m; i++ )
1.25 noro 1578: if ( b->body[i] ) {
1579: nd_free(b->body[i]);
1580: b->body[i] = 0;
1581: }
1582: GC_free(b);
1583: }
1584:
1.63 noro 1585: void add_pbucket_symbolic(PGeoBucket g,ND d)
1586: {
1587: int l,i,k,m;
1588:
1589: if ( !d )
1590: return;
1591: l = LEN(d);
1592: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1593: /* 2^(k-1) < l <= 2^k (=m) */
1594: d = nd_merge(g->body[k],d);
1595: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1596: g->body[k] = 0;
1597: d = nd_merge(g->body[k+1],d);
1598: }
1599: g->body[k] = d;
1600: g->m = MAX(g->m,k);
1601: }
1602:
1.31 noro 1603: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1604: {
1.31 noro 1605: int l,i,k,m;
1.1 noro 1606:
1.31 noro 1607: if ( !d )
1608: return;
1609: l = LEN(d);
1.29 noro 1610: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1611: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1612: d = nd_add(mod,g->body[k],d);
1613: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1614: g->body[k] = 0;
1.31 noro 1615: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1616: }
1617: g->body[k] = d;
1618: g->m = MAX(g->m,k);
1619: }
1620:
1.26 noro 1621: void mulq_pbucket(PGeoBucket g,Q c)
1622: {
1623: int k;
1624:
1625: for ( k = 0; k <= g->m; k++ )
1626: nd_mul_c_q(g->body[k],c);
1627: }
1628:
1.63 noro 1629: NM remove_head_pbucket_symbolic(PGeoBucket g)
1630: {
1631: int j,i,k,c;
1632: NM head;
1633:
1634: k = g->m;
1635: j = -1;
1636: for ( i = 0; i <= k; i++ ) {
1637: if ( !g->body[i] ) continue;
1638: if ( j < 0 ) j = i;
1639: else {
1640: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1641: if ( c > 0 )
1642: j = i;
1643: else if ( c == 0 )
1644: g->body[i] = nd_remove_head(g->body[i]);
1645: }
1646: }
1647: if ( j < 0 ) return 0;
1648: else {
1649: head = BDY(g->body[j]);
1650: if ( !NEXT(head) ) {
1651: FREEND(g->body[j]);
1652: g->body[j] = 0;
1653: } else {
1654: BDY(g->body[j]) = NEXT(head);
1655: LEN(g->body[j])--;
1656: }
1657: return head;
1658: }
1659: }
1660:
1.19 noro 1661: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1662: {
1663: int j,i,c,k,nv,sum;
1.61 noro 1664: UINT *di,*dj;
1.1 noro 1665: ND gi,gj;
1666:
1667: k = g->m;
1668: while ( 1 ) {
1669: j = -1;
1670: for ( i = 0; i <= k; i++ ) {
1671: if ( !(gi = g->body[i]) )
1672: continue;
1673: if ( j < 0 ) {
1674: j = i;
1675: gj = g->body[j];
1676: dj = HDL(gj);
1.14 noro 1677: sum = HCM(gj);
1.1 noro 1678: } else {
1.34 noro 1679: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1680: if ( c > 0 ) {
1.34 noro 1681: if ( sum ) HCM(gj) = sum;
1682: else g->body[j] = nd_remove_head(gj);
1.1 noro 1683: j = i;
1684: gj = g->body[j];
1685: dj = HDL(gj);
1.14 noro 1686: sum = HCM(gj);
1.1 noro 1687: } else if ( c == 0 ) {
1.71 noro 1688: if ( mod == -1 )
1689: sum = _addsf(sum,HCM(gi));
1690: else {
1691: sum = sum+HCM(gi)-mod;
1692: if ( sum < 0 ) sum += mod;
1693: }
1.1 noro 1694: g->body[i] = nd_remove_head(gi);
1695: }
1696: }
1697: }
1.34 noro 1698: if ( j < 0 ) return -1;
1.1 noro 1699: else if ( sum ) {
1.14 noro 1700: HCM(gj) = sum;
1.26 noro 1701: return j;
1.31 noro 1702: } else
1.26 noro 1703: g->body[j] = nd_remove_head(gj);
1704: }
1705: }
1706:
1707: int head_pbucket_q(PGeoBucket g)
1708: {
1709: int j,i,c,k,nv;
1710: Q sum,t;
1711: ND gi,gj;
1712:
1713: k = g->m;
1714: while ( 1 ) {
1715: j = -1;
1716: for ( i = 0; i <= k; i++ ) {
1.34 noro 1717: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1718: if ( j < 0 ) {
1719: j = i;
1720: gj = g->body[j];
1721: sum = HCQ(gj);
1722: } else {
1723: nv = NV(gi);
1.34 noro 1724: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1725: if ( c > 0 ) {
1.34 noro 1726: if ( sum ) HCQ(gj) = sum;
1727: else g->body[j] = nd_remove_head(gj);
1.26 noro 1728: j = i;
1729: gj = g->body[j];
1730: sum = HCQ(gj);
1731: } else if ( c == 0 ) {
1732: addq(sum,HCQ(gi),&t);
1733: sum = t;
1734: g->body[i] = nd_remove_head(gi);
1735: }
1736: }
1737: }
1.34 noro 1738: if ( j < 0 ) return -1;
1.26 noro 1739: else if ( sum ) {
1740: HCQ(gj) = sum;
1.1 noro 1741: return j;
1.31 noro 1742: } else
1.1 noro 1743: g->body[j] = nd_remove_head(gj);
1744: }
1745: }
1746:
1.25 noro 1747: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1748: {
1.31 noro 1749: int i;
1.1 noro 1750: ND r,t;
1751:
1752: r = 0;
1.28 noro 1753: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1754: r = nd_add(mod,r,g->body[i]);
1.28 noro 1755: g->body[i] = 0;
1756: }
1757: g->m = -1;
1.1 noro 1758: return r;
1759: }
1760:
1.27 noro 1761: /* return value = 0 => input is not a GB */
1762:
1.77 noro 1763: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1764: {
1765: int i,nh,sugar,stat;
1.23 noro 1766: NODE r,g,t;
1.1 noro 1767: ND_pairs d;
1768: ND_pairs l;
1.77 noro 1769: ND h,nf,s,head;
1.63 noro 1770: NDV nfv;
1.77 noro 1771: Q q;
1.69 noro 1772: union oNDC dn;
1.1 noro 1773:
1.23 noro 1774: g = 0; d = 0;
1775: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1776: d = update_pairs(d,g,i);
1777: g = update_base(g,i);
1778: }
1779: sugar = 0;
1780: while ( d ) {
1781: again:
1782: l = nd_minp(d,&d);
1.14 noro 1783: if ( SG(l) != sugar ) {
1.77 noro 1784: if ( ishomo ) {
1785: for ( i = nd_psn-1; SG(nd_ps[i]) == sugar; i-- ) {
1786: s = ndvtond(m,nd_ps[i]);
1787: s = nd_separate_head(s,&head);
1788: nd_nf(m,s,nd_ps,1,&dn,&nf);
1789: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1790: nf = nd_add(m,head,nf);
1791: ndv_free(nd_ps[i]);
1792: nd_removecont(m,nf);
1793: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1794: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1795: }
1796: }
1.14 noro 1797: sugar = SG(l);
1.71 noro 1798: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1799: }
1.53 noro 1800: stat = nd_sp(m,0,l,&h);
1.1 noro 1801: if ( !stat ) {
1802: NEXT(l) = d; d = l;
1.20 noro 1803: d = nd_reconstruct(m,0,d);
1.1 noro 1804: goto again;
1805: }
1.41 noro 1806: #if USE_GEOBUCKET
1.69 noro 1807: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1808: #else
1.69 noro 1809: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1810: #endif
1.1 noro 1811: if ( !stat ) {
1812: NEXT(l) = d; d = l;
1.20 noro 1813: d = nd_reconstruct(m,0,d);
1.1 noro 1814: goto again;
1815: } else if ( nf ) {
1.27 noro 1816: if ( checkonly ) return 0;
1.71 noro 1817: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1818: nd_removecont(m,nf);
1819: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1820: nh = ndv_newps(m,nfv,0);
1.1 noro 1821: d = update_pairs(d,g,nh);
1822: g = update_base(g,nh);
1823: FREENDP(l);
1824: } else {
1.71 noro 1825: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1826: FREENDP(l);
1827: }
1828: }
1.77 noro 1829: if ( nd_demand )
1830: for ( t = g; t; t = NEXT(t) )
1831: BDY(t) = (pointer)ndv_load((int)BDY(t));
1832: else
1833: for ( t = g; t; t = NEXT(t) )
1834: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1835: return g;
1836: }
1837:
1.23 noro 1838: NODE nd_gb_trace(int m)
1.20 noro 1839: {
1840: int i,nh,sugar,stat;
1.23 noro 1841: NODE r,g,t;
1.20 noro 1842: ND_pairs d;
1843: ND_pairs l;
1844: ND h,nf,nfq;
1.63 noro 1845: NDV nfv,nfqv;
1.69 noro 1846: union oNDC dn;
1.20 noro 1847:
1.23 noro 1848: g = 0; d = 0;
1849: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1850: d = update_pairs(d,g,i);
1851: g = update_base(g,i);
1852: }
1853: sugar = 0;
1854: while ( d ) {
1855: again:
1856: l = nd_minp(d,&d);
1857: if ( SG(l) != sugar ) {
1858: sugar = SG(l);
1.71 noro 1859: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1860: }
1.53 noro 1861: stat = nd_sp(m,0,l,&h);
1.20 noro 1862: if ( !stat ) {
1863: NEXT(l) = d; d = l;
1864: d = nd_reconstruct(m,1,d);
1865: goto again;
1866: }
1.41 noro 1867: #if USE_GEOBUCKET
1.53 noro 1868: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1869: #else
1.69 noro 1870: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1871: #endif
1.20 noro 1872: if ( !stat ) {
1873: NEXT(l) = d; d = l;
1874: d = nd_reconstruct(m,1,d);
1875: goto again;
1876: } else if ( nf ) {
1.77 noro 1877: if ( nd_demand ) {
1878: nfqv = ndv_load(nd_psn);
1879: nfq = ndvtond(0,nfqv);
1880: } else
1881: nfq = 0;
1882: if ( !nfq ) {
1883: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1884: NEXT(l) = d; d = l;
1885: d = nd_reconstruct(m,1,d);
1886: goto again;
1887: }
1.70 noro 1888: }
1.20 noro 1889: if ( nfq ) {
1.70 noro 1890: /* m|HC(nfq) => failure */
1.63 noro 1891: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1892:
1.71 noro 1893: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1894: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1895: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 noro 1896: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1897: d = update_pairs(d,g,nh);
1898: g = update_base(g,nh);
1899: } else {
1.71 noro 1900: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1901: }
1902: } else {
1.71 noro 1903: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1904: }
1905: FREENDP(l);
1906: }
1.77 noro 1907: if ( nd_demand )
1908: for ( t = g; t; t = NEXT(t) )
1909: BDY(t) = (pointer)ndv_load((int)BDY(t));
1910: else
1911: for ( t = g; t; t = NEXT(t) )
1912: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1913: return g;
1914: }
1915:
1.23 noro 1916: int ndv_compare(NDV *p1,NDV *p2)
1917: {
1.34 noro 1918: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1919: }
1920:
1921: int ndv_compare_rev(NDV *p1,NDV *p2)
1922: {
1.34 noro 1923: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1924: }
1925:
1.61 noro 1926: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1927: {
1.69 noro 1928: int i,n,stat;
1929: ND nf,g,head;
1.23 noro 1930: NODE t,a0,a;
1.69 noro 1931: union oNDC dn;
1932: NDV *w;
1933: Q q;
1.23 noro 1934:
1.69 noro 1935: n = length(f);
1936: #if 0
1937: w = (NDV *)ALLOCA(n*sizeof(NDV));
1938: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1939: qsort(w,n,sizeof(NDV),
1940: (int (*)(const void *,const void *))ndv_compare);
1941: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1942: #endif
1943: ndv_setup(m,0,f);
1944: for ( i = 0; i < n; ) {
1945: g = ndvtond(m,nd_ps[i]);
1946: g = nd_separate_head(g,&head);
1947: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1948: if ( !stat )
1.69 noro 1949: nd_reconstruct(m,0,0);
1950: else {
1.71 noro 1951: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1952: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1953: nf = nd_add(m,head,nf);
1954: ndv_free(nd_ps[i]);
1.24 noro 1955: nd_removecont(m,nf);
1.69 noro 1956: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1957: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1958: i++;
1.23 noro 1959: }
1960: }
1.71 noro 1961: if ( DP_Print ) { printf("\n"); }
1.23 noro 1962: for ( a0 = 0, i = 0; i < n; i++ ) {
1963: NEXTNODE(a0,a);
1.69 noro 1964: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1965: }
1966: NEXT(a) = 0;
1967: return a0;
1968: }
1969:
1.1 noro 1970: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1971: {
1972: ND_pairs d1,nd,cur,head,prev,remove;
1973:
1974: if ( !g ) return d;
1975: d = crit_B(d,t);
1976: d1 = nd_newpairs(g,t);
1977: d1 = crit_M(d1);
1978: d1 = crit_F(d1);
1.55 noro 1979: if ( do_weyl )
1980: head = d1;
1981: else {
1982: prev = 0; cur = head = d1;
1983: while ( cur ) {
1984: if ( crit_2( cur->i1,cur->i2 ) ) {
1985: remove = cur;
1986: if ( !prev ) head = cur = NEXT(cur);
1987: else cur = NEXT(prev) = NEXT(cur);
1988: FREENDP(remove);
1989: } else {
1990: prev = cur; cur = NEXT(cur);
1991: }
1.1 noro 1992: }
1993: }
1994: if ( !d )
1995: return head;
1996: else {
1997: nd = d;
1.34 noro 1998: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1999: NEXT(nd) = head;
2000: return d;
2001: }
2002: }
2003:
2004: ND_pairs nd_newpairs( NODE g, int t )
2005: {
2006: NODE h;
1.61 noro 2007: UINT *dl;
1.34 noro 2008: int ts,s;
1.1 noro 2009: ND_pairs r,r0;
2010:
1.20 noro 2011: dl = DL(nd_psh[t]);
1.34 noro 2012: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 2013: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
2014: NEXTND_pairs(r0,r);
2015: r->i1 = (int)BDY(h);
2016: r->i2 = t;
1.20 noro 2017: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 2018: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
2019: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 2020: }
2021: NEXT(r) = 0;
2022: return r0;
2023: }
2024:
2025: ND_pairs crit_B( ND_pairs d, int s )
2026: {
2027: ND_pairs cur,head,prev,remove;
1.61 noro 2028: UINT *t,*tl,*lcm;
1.1 noro 2029: int td,tdl;
2030:
2031: if ( !d ) return 0;
1.20 noro 2032: t = DL(nd_psh[s]);
1.1 noro 2033: prev = 0;
2034: head = cur = d;
1.61 noro 2035: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 2036: while ( cur ) {
2037: tl = cur->lcm;
2038: if ( ndl_reducible(tl,t)
1.20 noro 2039: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
2040: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 2041: remove = cur;
2042: if ( !prev ) {
2043: head = cur = NEXT(cur);
2044: } else {
2045: cur = NEXT(prev) = NEXT(cur);
2046: }
2047: FREENDP(remove);
2048: } else {
1.34 noro 2049: prev = cur; cur = NEXT(cur);
1.1 noro 2050: }
2051: }
2052: return head;
2053: }
2054:
2055: ND_pairs crit_M( ND_pairs d1 )
2056: {
2057: ND_pairs e,d2,d3,dd,p;
1.61 noro 2058: UINT *id,*jd;
1.1 noro 2059:
2060: for ( dd = 0, e = d1; e; e = d3 ) {
2061: if ( !(d2 = NEXT(e)) ) {
2062: NEXT(e) = dd;
2063: return e;
2064: }
1.34 noro 2065: id = LCM(e);
1.1 noro 2066: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2067: p = NEXT(d2);
2068: jd = LCM(d2);
2069: if ( ndl_equal(jd,id) )
2070: ;
2071: else if ( TD(jd) > TD(id) )
1.1 noro 2072: if ( ndl_reducible(jd,id) ) continue;
2073: else ;
1.34 noro 2074: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2075: NEXT(d2) = d3;
2076: d3 = d2;
2077: }
2078: NEXT(e) = dd;
2079: dd = e;
2080: continue;
2081: /**/
2082: delit: NEXT(d2) = d3;
2083: d3 = d2;
2084: for ( ; p; p = d2 ) {
2085: d2 = NEXT(p);
2086: NEXT(p) = d3;
2087: d3 = p;
2088: }
2089: FREENDP(e);
2090: }
2091: return dd;
2092: }
2093:
2094: ND_pairs crit_F( ND_pairs d1 )
2095: {
2096: ND_pairs rest, head,remove;
2097: ND_pairs last, p, r, w;
2098: int s;
2099:
2100: for ( head = last = 0, p = d1; NEXT(p); ) {
2101: r = w = equivalent_pairs(p,&rest);
1.14 noro 2102: s = SG(r);
1.1 noro 2103: w = NEXT(w);
2104: while ( w ) {
2105: if ( crit_2(w->i1,w->i2) ) {
2106: r = w;
2107: w = NEXT(w);
2108: while ( w ) {
2109: remove = w;
2110: w = NEXT(w);
2111: FREENDP(remove);
2112: }
2113: break;
1.14 noro 2114: } else if ( SG(w) < s ) {
1.1 noro 2115: FREENDP(r);
2116: r = w;
1.14 noro 2117: s = SG(r);
1.1 noro 2118: w = NEXT(w);
2119: } else {
2120: remove = w;
2121: w = NEXT(w);
2122: FREENDP(remove);
2123: }
2124: }
2125: if ( last ) NEXT(last) = r;
2126: else head = r;
2127: NEXT(last = r) = 0;
2128: p = rest;
2129: if ( !p ) return head;
2130: }
2131: if ( !last ) return p;
2132: NEXT(last) = p;
2133: return head;
2134: }
2135:
2136: int crit_2( int dp1, int dp2 )
2137: {
1.20 noro 2138: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2139: }
2140:
1.40 noro 2141: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2142: {
2143: ND_pairs w,p,r,s;
1.61 noro 2144: UINT *d;
1.1 noro 2145:
2146: w = d1;
1.34 noro 2147: d = LCM(w);
1.1 noro 2148: s = NEXT(w);
2149: NEXT(w) = 0;
2150: for ( r = 0; s; s = p ) {
2151: p = NEXT(s);
1.34 noro 2152: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2153: NEXT(s) = w; w = s;
1.1 noro 2154: } else {
1.39 noro 2155: NEXT(s) = r; r = s;
1.1 noro 2156: }
2157: }
2158: *prest = r;
2159: return w;
2160: }
2161:
2162: NODE update_base(NODE nd,int ndp)
2163: {
1.61 noro 2164: UINT *dl, *dln;
1.1 noro 2165: NODE last, p, head;
2166:
1.20 noro 2167: dl = DL(nd_psh[ndp]);
1.1 noro 2168: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2169: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2170: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2171: p = NEXT(p);
2172: if ( last ) NEXT(last) = p;
2173: } else {
2174: if ( !last ) head = p;
2175: p = NEXT(last = p);
2176: }
2177: }
2178: head = append_one(head,ndp);
2179: return head;
2180: }
2181:
2182: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2183: {
2184: ND_pairs m,ml,p,l;
1.61 noro 2185: UINT *lcm;
1.33 noro 2186: int s,td,len,tlen,c,c1;
1.1 noro 2187:
2188: if ( !(p = NEXT(m = d)) ) {
2189: *prest = p;
2190: NEXT(m) = 0;
2191: return m;
2192: }
1.14 noro 2193: s = SG(m);
1.77 noro 2194: if ( !NoSugar ) {
2195: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2196: if ( (SG(p) < s)
2197: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2198: ml = l; m = p; s = SG(m);
2199: }
2200: } else {
2201: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2202: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2203: ml = l; m = p; s = SG(m);
2204: }
2205: }
1.1 noro 2206: if ( !ml ) *prest = NEXT(m);
2207: else {
2208: NEXT(ml) = NEXT(m);
2209: *prest = d;
2210: }
2211: NEXT(m) = 0;
2212: return m;
2213: }
2214:
1.63 noro 2215: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2216: {
1.69 noro 2217: int msugar,i;
1.63 noro 2218: ND_pairs t,dm0,dm,dr0,dr;
2219:
2220: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2221: if ( SG(t) < msugar ) msugar = SG(t);
2222: dm0 = 0; dr0 = 0;
1.69 noro 2223: for ( i = 0, t = d; t; t = NEXT(t) )
2224: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2225: if ( dm0 ) NEXT(dm) = t;
2226: else dm0 = t;
2227: dm = t;
1.69 noro 2228: i++;
1.63 noro 2229: } else {
2230: if ( dr0 ) NEXT(dr) = t;
2231: else dr0 = t;
2232: dr = t;
2233: }
2234: NEXT(dm) = 0;
2235: if ( dr0 ) NEXT(dr) = 0;
2236: *prest = dr0;
2237: return dm0;
2238: }
2239:
1.77 noro 2240: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2241: {
1.3 noro 2242: int len;
1.13 noro 2243: RHist r;
1.20 noro 2244: NDV b;
1.3 noro 2245:
1.1 noro 2246: if ( nd_psn == nd_pslen ) {
2247: nd_pslen *= 2;
1.11 noro 2248: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2249: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2250: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2251: nd_bound = (UINT **)
2252: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2253: }
1.39 noro 2254: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2255: nd_ps[nd_psn] = a;
1.39 noro 2256: if ( aq ) {
1.63 noro 2257: nd_ps_trace[nd_psn] = aq;
2258: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2259: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2260: } else {
1.63 noro 2261: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2262: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2263: }
1.77 noro 2264: if ( nd_demand ) {
2265: if ( aq ) {
2266: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2267: nd_ps_trace[nd_psn] = 0;
2268: } else {
2269: ndv_save(nd_ps[nd_psn],nd_psn);
2270: nd_ps[nd_psn] = 0;
2271: }
2272: }
1.1 noro 2273: return nd_psn++;
2274: }
2275:
1.61 noro 2276: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2277: {
1.5 noro 2278: int i,j,td,len,max;
1.1 noro 2279: NODE s,s0,f0;
1.61 noro 2280: UINT *d;
1.13 noro 2281: RHist r;
1.69 noro 2282: NDV *w;
1.61 noro 2283: NDV a,am;
1.11 noro 2284:
2285: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2286:
1.69 noro 2287: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2288: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2289: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2290: qsort(w,nd_psn,sizeof(NDV),
2291: (int (*)(const void *,const void *))ndv_compare);
2292: nd_pslen = 2*nd_psn;
1.11 noro 2293: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2294: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2295: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2296: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2297:
1.11 noro 2298: if ( !nd_red )
1.13 noro 2299: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2300: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2301: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2302: if ( trace ) {
1.69 noro 2303: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2304: ndv_removecont(0,a);
2305: am = nd_ps[i] = ndv_dup(mod,a);
2306: ndv_mod(mod,am);
2307: ndv_removecont(mod,am);
2308: } else {
1.69 noro 2309: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2310: ndv_removecont(mod,a);
1.39 noro 2311: }
1.61 noro 2312: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2313: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2314: nd_psh[i] = r;
1.77 noro 2315: if ( nd_demand ) {
2316: if ( trace ) {
2317: ndv_save(nd_ps_trace[i],i);
2318: nd_ps_trace[i] = 0;
2319: } else {
2320: ndv_save(nd_ps[i],i);
2321: nd_ps[i] = 0;
2322: }
2323: }
1.20 noro 2324: }
2325: }
2326:
1.63 noro 2327: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2328: {
1.61 noro 2329: VL tv,fv,vv,vc;
1.1 noro 2330: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2331: int e,max,nvar;
1.69 noro 2332: NDV b;
1.77 noro 2333: int ishomo;
1.1 noro 2334:
1.78 noro 2335: if ( !m && Demand ) nd_demand = 1;
2336: else nd_demand = 0;
2337:
1.74 noro 2338: ndv_alloc = 0;
1.1 noro 2339: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2340: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.32 noro 2341: nd_init_ord(ord);
1.61 noro 2342: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2343: for ( tv = vv; tv; tv = NEXT(tv) ) {
2344: e = getdeg(tv->v,(P)BDY(t));
2345: max = MAX(e,max);
2346: }
2347: nd_setup_parameters(nvar,max);
1.77 noro 2348: ishomo = 1;
1.1 noro 2349: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2350: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2351: if ( ishomo )
2352: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2353: if ( m ) ndv_mod(m,b);
1.61 noro 2354: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2355: }
2356: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2357: ndv_setup(m,0,fd0);
1.77 noro 2358: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.61 noro 2359: x = ndv_reducebase(x);
2360: x = ndv_reduceall(m,x);
1.23 noro 2361: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2362: NEXTNODE(r0,r);
1.61 noro 2363: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2364: }
2365: if ( r0 ) NEXT(r) = 0;
2366: MKLIST(*rp,r0);
1.74 noro 2367: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2368: }
2369:
1.52 noro 2370: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2371: {
2372: struct order_spec ord1;
1.61 noro 2373: VL tv,fv,vv,vc;
1.27 noro 2374: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2375: int m,nocheck,nvar,mindex,e,max;
2376: NDV c;
2377: NMV a;
1.27 noro 2378: P p;
1.61 noro 2379: EPOS oepos;
2380: int obpe,oadv,wmax,i,len,cbpe;
1.20 noro 2381:
2382: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2383: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.52 noro 2384: nocheck = 0;
2385: mindex = 0;
2386:
1.78 noro 2387: if ( Demand ) nd_demand = 1;
2388: else nd_demand = 0;
2389:
1.52 noro 2390: /* setup modulus */
2391: if ( trace < 0 ) {
2392: trace = -trace;
2393: nocheck = 1;
2394: }
2395: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2396: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2397: for ( tv = vv; tv; tv = NEXT(tv) ) {
2398: e = getdeg(tv->v,(P)BDY(t));
2399: max = MAX(e,max);
1.23 noro 2400: }
1.61 noro 2401: nd_init_ord(ord);
2402: nd_setup_parameters(nvar,max);
2403: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2404: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2405: c = ptondv(CO,vv,(P)BDY(t));
2406: if ( c ) {
2407: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2408: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2409: }
1.61 noro 2410: }
2411: if ( in0 ) NEXT(in) = 0;
2412: if ( fd0 ) NEXT(fd) = 0;
2413: if ( homo ) {
2414: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2415: c = (NDV)BDY(t); len = LEN(c);
2416: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2417: wmax = MAX(TD(DL(a)),wmax);
2418: }
2419: homogenize_order(ord,nvar,&ord1);
2420: nd_init_ord(&ord1);
2421: nd_setup_parameters(nvar+1,wmax);
2422: for ( t = fd0; t; t = NEXT(t) )
2423: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2424: }
1.52 noro 2425: while ( 1 ) {
1.77 noro 2426: if ( Demand )
2427: nd_demand = 1;
1.61 noro 2428: ndv_setup(m,1,fd0);
1.27 noro 2429: cand = nd_gb_trace(m);
1.52 noro 2430: if ( !cand ) {
2431: /* failure */
1.61 noro 2432: if ( trace > 1 ) { *rp = 0; return; }
2433: else m = get_lprime(++mindex);
1.52 noro 2434: continue;
2435: }
1.27 noro 2436: if ( homo ) {
2437: /* dehomogenization */
1.61 noro 2438: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2439: nd_init_ord(ord);
1.61 noro 2440: nd_setup_parameters(nvar,0);
1.27 noro 2441: }
1.77 noro 2442: nd_demand = 0;
1.61 noro 2443: cand = ndv_reducebase(cand);
2444: cand = ndv_reduceall(0,cand);
1.78 noro 2445: cbpe = nd_bpe;
1.61 noro 2446: if ( nocheck )
2447: break;
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;
1.73 noro 2469: nd_setup_parameters(nd_nvar,0);
1.61 noro 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) ) {
1.71 noro 2550: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2551: else printf("+%d*",CM(m));
1.14 noro 2552: ndl_print(DL(m));
1.1 noro 2553: }
2554: printf("\n");
2555: }
2556: }
2557:
1.16 noro 2558: void nd_print_q(ND p)
2559: {
2560: NM m;
2561:
2562: if ( !p )
2563: printf("0\n");
2564: else {
2565: for ( m = BDY(p); m; m = NEXT(m) ) {
2566: printf("+");
1.74 noro 2567: printexpr(CO,(Obj)CQ(m));
1.16 noro 2568: printf("*");
2569: ndl_print(DL(m));
2570: }
2571: printf("\n");
2572: }
2573: }
2574:
1.1 noro 2575: void ndp_print(ND_pairs d)
2576: {
2577: ND_pairs t;
2578:
1.34 noro 2579: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2580: printf("\n");
2581: }
2582:
1.20 noro 2583: void nd_removecont(int mod,ND p)
1.16 noro 2584: {
2585: int i,n;
2586: Q *w;
2587: Q dvr,t;
2588: NM m;
1.21 noro 2589: struct oVECT v;
2590: N q,r;
1.16 noro 2591:
1.71 noro 2592: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2593: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2594: else {
2595: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2596: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2597: v.len = n;
2598: v.body = (pointer *)w;
1.34 noro 2599: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2600: removecont_array(w,n);
2601: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2602: }
2603: }
2604:
1.21 noro 2605: void nd_removecont2(ND p1,ND p2)
2606: {
2607: int i,n1,n2,n;
2608: Q *w;
2609: Q dvr,t;
2610: NM m;
2611: struct oVECT v;
2612: N q,r;
2613:
2614: if ( !p1 ) {
2615: nd_removecont(0,p2); return;
2616: } else if ( !p2 ) {
2617: nd_removecont(0,p1); return;
2618: }
2619: n1 = nd_length(p1);
2620: n2 = nd_length(p2);
2621: n = n1+n2;
2622: w = (Q *)ALLOCA(n*sizeof(Q));
2623: v.len = n;
2624: v.body = (pointer *)w;
1.34 noro 2625: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2626: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2627: removecont_array(w,n);
2628: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2629: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2630: }
2631:
1.20 noro 2632: void ndv_removecont(int mod,NDV p)
1.16 noro 2633: {
2634: int i,len;
2635: Q *w;
2636: Q dvr,t;
2637: NMV m;
2638:
1.71 noro 2639: if ( mod == -1 )
2640: ndv_mul_c(mod,p,_invsf(HCM(p)));
2641: else if ( mod )
1.20 noro 2642: ndv_mul_c(mod,p,invm(HCM(p),mod));
2643: else {
2644: len = p->len;
2645: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2646: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2647: sortbynm(w,len);
2648: qltozl(w,len,&dvr);
2649: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2650: divq(CQ(m),dvr,&t); CQ(m) = t;
2651: }
1.16 noro 2652: }
1.21 noro 2653: }
2654:
1.61 noro 2655: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2656: {
2657: int len,i,max;
2658: NMV m,mr0,mr,t;
2659:
2660: len = p->len;
2661: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2662: max = MAX(max,TD(DL(m)));
2663: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2664: m = (NMV)((char *)mr0+(len-1)*oadv);
2665: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2666: t = (NMV)ALLOCA(nmv_adv);
2667: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2668: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2669: CQ(mr) = CQ(m);
2670: ndl_copy(DL(t),DL(mr));
2671: }
2672: NV(p)++;
2673: BDY(p) = mr0;
2674: }
2675:
1.45 noro 2676: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2677: {
1.45 noro 2678: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2679: Q *w;
2680: Q dvr,t;
2681: NMV m,r;
2682:
2683: len = p->len;
2684: newnvar = nd_nvar-1;
1.48 noro 2685: newexporigin = nd_get_exporigin(ord);
1.45 noro 2686: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2687: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2688: ndl_dehomogenize(DL(m));
1.23 noro 2689: if ( newwpd != nd_wpd ) {
1.61 noro 2690: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2691: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2692: CQ(r) = CQ(m);
2693: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2694: adj = nd_exporigin-newexporigin;
2695: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2696: }
2697: }
2698: NV(p)--;
2699: }
2700:
1.21 noro 2701: void removecont_array(Q *c,int n)
2702: {
2703: struct oVECT v;
2704: Q d0,d1,a,u,u1,gcd;
2705: int i;
2706: N qn,rn,gn;
2707: Q *q,*r;
2708:
2709: q = (Q *)ALLOCA(n*sizeof(Q));
2710: r = (Q *)ALLOCA(n*sizeof(Q));
2711: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2712: igcdv_estimate(&v,&d0);
2713: for ( i = 0; i < n; i++ ) {
2714: divn(NM(c[i]),NM(d0),&qn,&rn);
2715: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2716: NTOQ(rn,SGN(c[i]),r[i]);
2717: }
1.34 noro 2718: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2719: if ( i < n ) {
2720: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2721: igcdv(&v,&d1);
2722: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2723: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2724: for ( i = 0; i < n; i++ ) {
2725: mulq(a,q[i],&u);
2726: if ( r[i] ) {
2727: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2728: addq(u,u1,&q[i]);
2729: } else
2730: q[i] = u;
2731: }
2732: }
1.34 noro 2733: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2734: }
2735:
1.19 noro 2736: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2737: {
2738: NM m;
2739: int c,c1;
2740:
1.34 noro 2741: if ( !p ) return;
1.71 noro 2742: if ( mod == -1 )
2743: for ( m = BDY(p); m; m = NEXT(m) )
2744: CM(m) = _mulsf(CM(m),mul);
2745: else
2746: for ( m = BDY(p); m; m = NEXT(m) ) {
2747: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2748: }
1.1 noro 2749: }
2750:
1.16 noro 2751: void nd_mul_c_q(ND p,Q mul)
2752: {
2753: NM m;
2754: Q c;
2755:
1.34 noro 2756: if ( !p ) return;
1.16 noro 2757: for ( m = BDY(p); m; m = NEXT(m) ) {
2758: mulq(CQ(m),mul,&c); CQ(m) = c;
2759: }
2760: }
2761:
1.61 noro 2762: void nd_mul_c_p(VL vl,ND p,P mul)
2763: {
2764: NM m;
2765: P c;
2766:
2767: if ( !p ) return;
2768: for ( m = BDY(p); m; m = NEXT(m) ) {
2769: mulp(vl,CP(m),mul,&c); CP(m) = c;
2770: }
2771: }
2772:
1.1 noro 2773: void nd_free(ND p)
2774: {
2775: NM t,s;
2776:
1.34 noro 2777: if ( !p ) return;
1.1 noro 2778: t = BDY(p);
2779: while ( t ) {
2780: s = NEXT(t);
2781: FREENM(t);
2782: t = s;
2783: }
2784: FREEND(p);
2785: }
2786:
1.23 noro 2787: void ndv_free(NDV p)
2788: {
2789: GC_free(BDY(p));
2790: }
2791:
1.61 noro 2792: void nd_append_red(UINT *d,int i)
1.1 noro 2793: {
1.13 noro 2794: RHist m,m0;
1.1 noro 2795: int h;
2796:
1.13 noro 2797: NEWRHist(m);
1.34 noro 2798: h = ndl_hash_value(d);
1.13 noro 2799: m->index = i;
1.14 noro 2800: ndl_copy(d,DL(m));
1.1 noro 2801: NEXT(m) = nd_red[h];
2802: nd_red[h] = m;
2803: }
2804:
1.61 noro 2805: UINT *ndv_compute_bound(NDV p)
1.1 noro 2806: {
1.61 noro 2807: UINT *d1,*d2,*t;
2808: UINT u;
1.57 noro 2809: int i,j,k,l,len,ind;
1.45 noro 2810: NMV m;
1.1 noro 2811:
2812: if ( !p )
2813: return 0;
1.61 noro 2814: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2815: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2816: len = LEN(p);
2817: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2818: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2819: ndl_lcm(DL(m),d1,d2);
1.1 noro 2820: t = d1; d1 = d2; d2 = t;
2821: }
1.12 noro 2822: l = nd_nvar+31;
1.61 noro 2823: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2824: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2825: u = d1[i];
2826: k = (nd_epw-1)*nd_bpe;
2827: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2828: t[ind] = (u>>k)&nd_mask0;
2829: }
2830: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2831: return t;
2832: }
2833:
1.48 noro 2834: int nd_get_exporigin(struct order_spec *ord)
2835: {
1.51 noro 2836: switch ( ord->id ) {
1.41 noro 2837: case 0:
1.48 noro 2838: return 1;
1.41 noro 2839: case 1:
2840: /* block order */
1.43 noro 2841: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2842: return ord->ord.block.length+1;
1.41 noro 2843: case 2:
1.52 noro 2844: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2845: }
1.48 noro 2846: }
2847:
1.61 noro 2848: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2849: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2850: struct order_pair *op;
1.48 noro 2851:
1.73 noro 2852: nd_nvar = nvar;
1.79 noro 2853: if ( max ) {
2854: if ( max < 2 ) nd_bpe = 1;
2855: else if ( max < 4 ) nd_bpe = 2;
2856: else if ( max < 8 ) nd_bpe = 3;
2857: else if ( max < 16 ) nd_bpe = 4;
2858: else if ( max < 32 ) nd_bpe = 5;
2859: else if ( max < 64 ) nd_bpe = 6;
2860: else if ( max < 256 ) nd_bpe = 8;
2861: else if ( max < 1024 ) nd_bpe = 10;
2862: else if ( max < 65536 ) nd_bpe = 16;
2863: else nd_bpe = 32;
2864: }
1.61 noro 2865: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2866: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2867: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2868: wpd = nd_exporigin+elen;
2869: if ( wpd != nd_wpd ) {
2870: nd_free_private_storage();
2871: nd_wpd = wpd;
2872: }
1.1 noro 2873: if ( nd_bpe < 32 ) {
2874: nd_mask0 = (1<<nd_bpe)-1;
2875: } else {
2876: nd_mask0 = 0xffffffff;
2877: }
2878: bzero(nd_mask,sizeof(nd_mask));
2879: nd_mask1 = 0;
2880: for ( i = 0; i < nd_epw; i++ ) {
2881: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2882: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2883: }
1.61 noro 2884: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2885: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2886: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2887: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2888: }
2889:
1.20 noro 2890: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2891: {
1.37 noro 2892: int i,obpe,oadv,h;
1.71 noro 2893: static NM prev_nm_free_list;
2894: static ND_pairs prev_ndp_free_list;
1.13 noro 2895: RHist mr0,mr;
2896: RHist r;
1.37 noro 2897: RHist *old_red;
1.71 noro 2898: ND_pairs s0,s,t;
1.43 noro 2899: EPOS oepos;
1.15 noro 2900:
1.1 noro 2901: obpe = nd_bpe;
1.11 noro 2902: oadv = nmv_adv;
1.43 noro 2903: oepos = nd_epos;
1.63 noro 2904: if ( obpe < 2 ) nd_bpe = 2;
2905: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2906: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2907: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2908: else if ( obpe < 6 ) nd_bpe = 6;
2909: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2910: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2911: else if ( obpe < 16 ) nd_bpe = 16;
2912: else if ( obpe < 32 ) nd_bpe = 32;
2913: else error("nd_reconstruct : exponent too large");
1.5 noro 2914:
1.73 noro 2915: nd_setup_parameters(nd_nvar,0);
1.1 noro 2916: prev_nm_free_list = _nm_free_list;
2917: prev_ndp_free_list = _ndp_free_list;
2918: _nm_free_list = 0;
2919: _ndp_free_list = 0;
1.53 noro 2920: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2921: if ( trace )
2922: for ( i = nd_psn-1; i >= 0; i-- )
2923: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2924: s0 = 0;
2925: for ( t = d; t; t = NEXT(t) ) {
2926: NEXTND_pairs(s0,s);
2927: s->i1 = t->i1;
2928: s->i2 = t->i2;
1.14 noro 2929: SG(s) = SG(t);
1.61 noro 2930: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2931: }
1.37 noro 2932:
2933: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2934: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2935: old_red[i] = nd_red[i];
2936: nd_red[i] = 0;
2937: }
2938: for ( i = 0; i < REDTAB_LEN; i++ )
2939: for ( r = old_red[i]; r; r = NEXT(r) ) {
2940: NEWRHist(mr);
1.13 noro 2941: mr->index = r->index;
1.20 noro 2942: SG(mr) = SG(r);
1.61 noro 2943: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2944: h = ndl_hash_value(DL(mr));
2945: NEXT(mr) = nd_red[h];
2946: nd_red[h] = mr;
1.6 noro 2947: }
1.37 noro 2948: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2949: old_red = 0;
1.11 noro 2950: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2951: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2952: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2953: nd_psh[i] = r;
1.11 noro 2954: }
1.1 noro 2955: if ( s0 ) NEXT(s) = 0;
2956: prev_nm_free_list = 0;
2957: prev_ndp_free_list = 0;
1.71 noro 2958: #if 0
1.1 noro 2959: GC_gcollect();
1.71 noro 2960: #endif
1.1 noro 2961: return s0;
2962: }
2963:
1.61 noro 2964: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2965: {
1.57 noro 2966: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2967: struct order_pair *op;
1.1 noro 2968:
2969: n = nd_nvar;
1.61 noro 2970: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2971: omask0 = (1<<obpe)-1;
1.34 noro 2972: TD(r) = TD(d);
1.41 noro 2973: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2974: if ( nd_blockmask ) {
2975: l = nd_blockmask->n;
2976: op = nd_blockmask->order_pair;
2977: for ( i = 1; i < nd_exporigin; i++ )
2978: r[i] = d[i];
2979: for ( j = 0, s = 0; j < l; j++ ) {
2980: ord_l = op[j].length;
1.57 noro 2981: for ( i = 0; i < ord_l; i++, s++ ) {
2982: ei = GET_EXP_OLD(d,s);
2983: PUT_EXP(r,s,ei);
2984: }
1.1 noro 2985: }
1.43 noro 2986: } else {
1.56 noro 2987: for ( i = 0; i < n; i++ ) {
2988: ei = GET_EXP_OLD(d,i);
2989: PUT_EXP(r,i,ei);
2990: }
1.1 noro 2991: }
2992: }
1.3 noro 2993:
1.6 noro 2994: ND nd_copy(ND p)
2995: {
2996: NM m,mr,mr0;
1.41 noro 2997: int c,n;
1.6 noro 2998: ND r;
2999:
3000: if ( !p )
3001: return 0;
3002: else {
3003: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3004: NEXTNM(mr0,mr);
1.14 noro 3005: CM(mr) = CM(m);
3006: ndl_copy(DL(m),DL(mr));
1.6 noro 3007: }
3008: NEXT(mr) = 0;
1.31 noro 3009: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3010: SG(r) = SG(p);
1.6 noro 3011: return r;
3012: }
3013: }
3014:
1.53 noro 3015: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3016: {
3017: NM m;
3018: NDV p1,p2;
3019: ND t1,t2;
1.61 noro 3020: UINT *lcm;
1.31 noro 3021: int td;
1.11 noro 3022:
1.77 noro 3023: if ( !mod && nd_demand ) {
3024: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3025: } else {
1.77 noro 3026: if ( trace ) {
3027: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3028: } else {
3029: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3030: }
1.20 noro 3031: }
1.34 noro 3032: lcm = LCM(p);
1.11 noro 3033: NEWNM(m);
1.20 noro 3034: CQ(m) = HCQ(p2);
1.34 noro 3035: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3036: if ( ndl_check_bound2(p->i1,DL(m)) )
3037: return 0;
1.55 noro 3038: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3039: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3040: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 3041: else chsgnq(HCQ(p1),&CQ(m));
3042: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3043: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3044: nd_free(t1);
3045: return 0;
3046: }
1.55 noro 3047: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3048: *rp = nd_add(mod,t1,t2);
1.11 noro 3049: FREENM(m);
3050: return 1;
3051: }
3052:
1.19 noro 3053: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3054: {
3055: NMV m;
3056: int c,c1,len,i;
3057:
1.34 noro 3058: if ( !p ) return;
1.14 noro 3059: len = LEN(p);
1.71 noro 3060: if ( mod == -1 )
3061: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3062: CM(m) = _mulsf(CM(m),mul);
3063: else
3064: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3065: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3066: }
1.11 noro 3067: }
3068:
1.16 noro 3069: void ndv_mul_c_q(NDV p,Q mul)
3070: {
3071: NMV m;
3072: Q c;
3073: int len,i;
3074:
1.34 noro 3075: if ( !p ) return;
1.16 noro 3076: len = LEN(p);
3077: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3078: mulq(CQ(m),mul,&c); CQ(m) = c;
3079: }
3080: }
3081:
1.55 noro 3082: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3083: int n2,i,j,l,n,tlen;
1.61 noro 3084: UINT *d0;
1.55 noro 3085: NM *tab,*psum;
3086: ND s,r;
3087: NM t;
3088: NMV m1;
3089:
3090: if ( !p ) return 0;
3091: n = NV(p); n2 = n>>1;
3092: d0 = DL(m0);
3093: l = LEN(p);
3094: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3095: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3096: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3097: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3098: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3099: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3100: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3101: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3102: for ( j = 0; j < tlen; j++ ) {
3103: if ( tab[j] ) {
3104: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3105: }
3106: }
3107: }
3108: for ( i = tlen-1, r = 0; i >= 0; i-- )
3109: if ( psum[i] ) {
3110: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3111: MKND(n,psum[i],j,s);
3112: r = nd_add(mod,r,s);
3113: }
1.56 noro 3114: if ( r ) SG(r) = SG(p)+TD(d0);
3115: return r;
1.55 noro 3116: }
3117:
1.56 noro 3118: /* product of monomials */
3119: /* XXX block order is not handled correctly */
3120:
1.55 noro 3121: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3122: {
1.56 noro 3123: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3124: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3125: Q *ctab_q;
3126: Q q,q1;
1.61 noro 3127: UINT c0,c1,c;
1.55 noro 3128: NM *p;
3129: NM m,t;
3130:
3131: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3132: if ( !m0 || !m1 ) return;
3133: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3134: NEWNM(m); d = DL(m);
1.56 noro 3135: if ( mod ) {
3136: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3137: } else
3138: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3139: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3140: homo = n&1 ? 1 : 0;
3141: if ( homo ) {
3142: /* offset of h-degree */
3143: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3144: PUT_EXP(DL(m),n-1,h);
3145: TD(DL(m)) = h;
1.61 noro 3146: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3147: }
3148: tab[0] = m;
3149: NEWNM(m); d = DL(m);
1.57 noro 3150: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3151: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3152: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3153: /* xi^a*(Di^k*xi^l)*Di^b */
3154: a += l; b += k;
1.56 noro 3155: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3156: if ( !k || !l ) {
3157: for ( j = 0; j < curlen; j++ )
1.56 noro 3158: if ( t = tab[j] ) {
3159: dt = DL(t);
3160: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3161: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3162: }
3163: curlen *= k+1;
3164: continue;
3165: }
3166: min = MIN(k,l);
1.56 noro 3167: if ( mod ) {
1.61 noro 3168: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3169: mkwcm(k,l,mod,ctab);
3170: } else {
3171: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3172: mkwc(k,l,ctab_q);
3173: }
1.57 noro 3174: for ( j = min; j >= 0; j-- ) {
1.56 noro 3175: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3176: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3177: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3178: if ( homo ) {
3179: TD(d) = s;
1.56 noro 3180: PUT_EXP(d,n-1,s-h);
1.55 noro 3181: } else TD(d) = h;
1.61 noro 3182: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3183: if ( mod ) c = ctab[j];
3184: else q = ctab_q[j];
1.57 noro 3185: p = tab+curlen*j;
3186: if ( j == 0 ) {
3187: for ( u = 0; u < curlen; u++, p++ ) {
3188: if ( tab[u] ) {
3189: ndl_addto(DL(tab[u]),d);
3190: if ( mod ) {
3191: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3192: } else {
3193: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3194: }
3195: }
1.56 noro 3196: }
1.57 noro 3197: } else {
3198: for ( u = 0; u < curlen; u++, p++ ) {
3199: if ( tab[u] ) {
3200: NEWNM(t);
3201: ndl_add(DL(tab[u]),d,DL(t));
3202: if ( mod ) {
3203: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3204: } else
3205: mulq(CQ(tab[u]),q,&CQ(t));
3206: *p = t;
3207: }
1.55 noro 3208: }
3209: }
3210: }
3211: curlen *= k+1;
3212: }
3213: FREENM(m);
3214: }
3215:
1.63 noro 3216: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3217: {
3218: NM mr,mr0;
3219: NMV m;
3220: UINT *d,*dt,*dm;
3221: int c,n,td,i,c1,c2,len;
3222: Q q;
3223: ND r;
3224:
3225: if ( !p ) return 0;
3226: else {
3227: n = NV(p); m = BDY(p);
3228: d = DL(m0);
3229: len = LEN(p);
3230: mr0 = 0;
3231: td = TD(d);
3232: c = CM(m0);
3233: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3234: NEXTNM(mr0,mr);
3235: CM(mr) = 1;
3236: ndl_add(DL(m),d,DL(mr));
3237: }
3238: NEXT(mr) = 0;
3239: MKND(NV(p),mr0,len,r);
3240: SG(r) = SG(p) + TD(d);
3241: return r;
3242: }
3243: }
3244:
1.55 noro 3245: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3246: {
3247: NM mr,mr0;
3248: NMV m;
1.61 noro 3249: UINT *d,*dt,*dm;
1.9 noro 3250: int c,n,td,i,c1,c2,len;
1.16 noro 3251: Q q;
1.9 noro 3252: ND r;
3253:
1.34 noro 3254: if ( !p ) return 0;
1.55 noro 3255: else if ( do_weyl )
1.71 noro 3256: if ( mod == -1 )
3257: error("ndv_mul_nm : not implemented (weyl)");
3258: else
3259: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3260: else {
3261: n = NV(p); m = BDY(p);
1.34 noro 3262: d = DL(m0);
1.14 noro 3263: len = LEN(p);
1.9 noro 3264: mr0 = 0;
1.34 noro 3265: td = TD(d);
1.71 noro 3266: if ( mod == -1 ) {
3267: c = CM(m0);
3268: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3269: NEXTNM(mr0,mr);
3270: CM(mr) = _mulsf(CM(m),c);
3271: ndl_add(DL(m),d,DL(mr));
3272: }
3273: } else if ( mod ) {
1.16 noro 3274: c = CM(m0);
3275: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3276: NEXTNM(mr0,mr);
3277: c1 = CM(m);
1.19 noro 3278: DMAR(c1,c,0,mod,c2);
1.16 noro 3279: CM(mr) = c2;
3280: ndl_add(DL(m),d,DL(mr));
3281: }
3282: } else {
3283: q = CQ(m0);
3284: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3285: NEXTNM(mr0,mr);
3286: mulq(CQ(m),q,&CQ(mr));
3287: ndl_add(DL(m),d,DL(mr));
3288: }
1.4 noro 3289: }
1.9 noro 3290: NEXT(mr) = 0;
1.31 noro 3291: MKND(NV(p),mr0,len,r);
1.34 noro 3292: SG(r) = SG(p) + TD(d);
1.9 noro 3293: return r;
1.4 noro 3294: }
3295: }
3296:
1.43 noro 3297: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3298: {
1.13 noro 3299: NMV m,mr,mr0,t;
3300: int len,i,k;
1.11 noro 3301:
1.61 noro 3302: if ( !p ) return;
3303: m = BDY(p); len = LEN(p);
3304: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3305: m = (NMV)((char *)mr0+(len-1)*oadv);
3306: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3307: t = (NMV)ALLOCA(nmv_adv);
3308: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3309: CQ(t) = CQ(m);
3310: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3311: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3312: CQ(mr) = CQ(t);
3313: ndl_copy(DL(t),DL(mr));
3314: }
3315: BDY(p) = mr0;
3316: }
3317:
3318: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3319: {
3320: NMV m,mr,mr0;
3321: int len,i;
3322: NDV r;
1.11 noro 3323:
1.61 noro 3324: if ( !p ) return 0;
3325: m = BDY(p); len = LEN(p);
3326: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3327: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3328: ndl_zero(DL(mr));
3329: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3330: CQ(mr) = CQ(m);
1.11 noro 3331: }
1.61 noro 3332: MKNDV(NV(p),mr0,len,r);
3333: SG(r) = SG(p);
3334: return r;
1.11 noro 3335: }
3336:
1.61 noro 3337: /* duplicate p */
3338:
3339: NDV ndv_dup(int mod,NDV p)
1.3 noro 3340: {
3341: NDV d;
1.61 noro 3342: NMV t,m,m0;
1.3 noro 3343: int i,len;
3344:
1.34 noro 3345: if ( !p ) return 0;
1.31 noro 3346: len = LEN(p);
1.34 noro 3347: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3348: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3349: ndl_copy(DL(t),DL(m));
1.16 noro 3350: CQ(m) = CQ(t);
1.3 noro 3351: }
3352: MKNDV(NV(p),m0,len,d);
1.23 noro 3353: SG(d) = SG(p);
3354: return d;
3355: }
3356:
1.63 noro 3357: ND nd_dup(ND p)
3358: {
3359: ND d;
3360: NM t,m,m0;
3361:
3362: if ( !p ) return 0;
3363: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3364: NEXTNM(m0,m);
3365: ndl_copy(DL(t),DL(m));
3366: CQ(m) = CQ(t);
3367: }
3368: if ( m0 ) NEXT(m) = 0;
3369: MKND(NV(p),m0,LEN(p),d);
3370: SG(d) = SG(p);
3371: return d;
3372: }
3373:
1.61 noro 3374: /* XXX if p->len == 0 then it represents 0 */
3375:
3376: void ndv_mod(int mod,NDV p)
3377: {
3378: NMV t,d;
3379: int r;
3380: int i,len,dlen;
1.71 noro 3381: Obj gfs;
1.61 noro 3382:
3383: if ( !p ) return;
3384: len = LEN(p);
3385: dlen = 0;
1.71 noro 3386: if ( mod == -1 )
3387: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3388: simp_ff((Obj)CP(t),&gfs);
3389: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3390: CM(d) = r;
3391: ndl_copy(DL(t),DL(d));
3392: NMV_ADV(d);
3393: dlen++;
3394: }
1.71 noro 3395: else
3396: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3397: r = rem(NM(CQ(t)),mod);
3398: if ( r ) {
3399: if ( SGN(CQ(t)) < 0 )
3400: r = mod-r;
3401: CM(d) = r;
3402: ndl_copy(DL(t),DL(d));
3403: NMV_ADV(d);
3404: dlen++;
3405: }
3406: }
1.61 noro 3407: LEN(p) = dlen;
3408: }
3409:
3410: NDV ptondv(VL vl,VL dvl,P p)
3411: {
3412: ND nd;
3413:
3414: nd = ptond(vl,dvl,p);
3415: return ndtondv(0,nd);
3416: }
3417:
3418: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3419: {
1.61 noro 3420: int n,i,j,k,e;
3421: VL tvl;
3422: V v;
3423: DCP dc;
3424: DCP *w;
3425: ND r,s,t,u;
3426: P x;
3427: int c;
3428: UINT *d;
1.23 noro 3429: NM m,m0;
1.61 noro 3430:
3431: if ( !p )
3432: return 0;
3433: else if ( NUM(p) ) {
3434: NEWNM(m);
3435: ndl_zero(DL(m));
3436: CQ(m) = (Q)p;
3437: NEXT(m) = 0;
3438: MKND(nd_nvar,m,1,r);
3439: SG(r) = 0;
3440: return r;
3441: } else {
3442: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3443: w = (DCP *)ALLOCA(k*sizeof(DCP));
3444: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3445: for ( i = 0, tvl = dvl, v = VR(p);
3446: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3447: if ( !tvl ) {
3448: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3449: t = ptond(vl,dvl,COEF(w[j]));
3450: pwrp(vl,x,DEG(w[j]),&p);
3451: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3452: }
3453: return s;
3454: } else {
3455: NEWNM(m0); d = DL(m0);
3456: for ( j = k-1, s = 0; j >= 0; j-- ) {
3457: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3458: TD(d) = MUL_WEIGHT(e,i);
3459: if ( nd_blockmask) ndl_weight_mask(d);
3460: t = ptond(vl,dvl,COEF(w[j]));
3461: for ( m = BDY(t); m; m = NEXT(m) )
3462: ndl_addto(DL(m),d);
3463: SG(t) += TD(d);
3464: s = nd_add(0,s,t);
3465: }
3466: FREENM(m0);
3467: return s;
3468: }
3469: }
3470: }
3471:
3472: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3473: {
3474: VL tvl;
3475: int len,n,j,i,e;
3476: NMV m;
3477: Q q;
3478: P c;
3479: UINT *d;
3480: P s,r,u,t,w;
1.71 noro 3481: GFS gfs;
1.23 noro 3482:
1.34 noro 3483: if ( !p ) return 0;
1.61 noro 3484: else {
3485: len = LEN(p);
3486: n = NV(p);
3487: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3488: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3489: if ( mod == -1 ) {
3490: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3491: } else if ( mod ) {
1.61 noro 3492: STOQ(CM(m),q); c = (P)q;
3493: } else
3494: c = CP(m);
3495: d = DL(m);
3496: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3497: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3498: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3499: }
3500: addp(vl,s,t,&u); s = u;
3501: }
3502: return s;
1.23 noro 3503: }
1.3 noro 3504: }
3505:
1.61 noro 3506: NDV ndtondv(int mod,ND p)
1.11 noro 3507: {
3508: NDV d;
1.61 noro 3509: NMV m,m0;
3510: NM t;
3511: int i,len;
1.11 noro 3512:
1.34 noro 3513: if ( !p ) return 0;
1.61 noro 3514: len = LEN(p);
1.74 noro 3515: if ( mod )
3516: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3517: else
3518: m0 = m = MALLOC(len*nmv_adv);
3519: ndv_alloc += nmv_adv*len;
1.61 noro 3520: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3521: ndl_copy(DL(t),DL(m));
3522: CQ(m) = CQ(t);
1.11 noro 3523: }
1.61 noro 3524: MKNDV(NV(p),m0,len,d);
1.14 noro 3525: SG(d) = SG(p);
1.11 noro 3526: return d;
3527: }
3528:
1.61 noro 3529: ND ndvtond(int mod,NDV p)
1.11 noro 3530: {
1.61 noro 3531: ND d;
3532: NM m,m0;
1.11 noro 3533: NMV t;
1.61 noro 3534: int i,len;
1.11 noro 3535:
1.34 noro 3536: if ( !p ) return 0;
1.11 noro 3537: m0 = 0;
1.61 noro 3538: len = p->len;
3539: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3540: NEXTNM(m0,m);
3541: ndl_copy(DL(t),DL(m));
3542: CQ(m) = CQ(t);
1.11 noro 3543: }
3544: NEXT(m) = 0;
1.61 noro 3545: MKND(NV(p),m0,len,d);
1.14 noro 3546: SG(d) = SG(p);
1.11 noro 3547: return d;
3548: }
3549:
1.3 noro 3550: void ndv_print(NDV p)
3551: {
3552: NMV m;
3553: int i,len;
3554:
1.34 noro 3555: if ( !p ) printf("0\n");
1.3 noro 3556: else {
1.14 noro 3557: len = LEN(p);
1.3 noro 3558: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3559: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3560: else printf("+%d*",CM(m));
1.16 noro 3561: ndl_print(DL(m));
3562: }
3563: printf("\n");
3564: }
3565: }
3566:
3567: void ndv_print_q(NDV p)
3568: {
3569: NMV m;
3570: int i,len;
3571:
1.34 noro 3572: if ( !p ) printf("0\n");
1.16 noro 3573: else {
3574: len = LEN(p);
3575: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3576: printf("+");
1.74 noro 3577: printexpr(CO,(Obj)CQ(m));
1.16 noro 3578: printf("*");
1.14 noro 3579: ndl_print(DL(m));
1.3 noro 3580: }
3581: printf("\n");
3582: }
1.25 noro 3583: }
3584:
1.61 noro 3585: NODE ndv_reducebase(NODE x)
1.27 noro 3586: {
3587: int len,i,j;
3588: NDV *w;
3589: NODE t,t0;
3590:
3591: len = length(x);
3592: w = (NDV *)ALLOCA(len*sizeof(NDV));
3593: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3594: for ( i = 0; i < len; i++ ) {
3595: for ( j = 0; j < i; j++ ) {
3596: if ( w[i] && w[j] )
3597: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3598: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3599: }
3600: }
3601: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3602: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3603: }
3604: NEXT(t) = 0; x = t0;
3605: return x;
1.11 noro 3606: }
1.32 noro 3607:
1.43 noro 3608: /* XXX incomplete */
3609:
1.32 noro 3610: void nd_init_ord(struct order_spec *ord)
3611: {
1.43 noro 3612: switch ( ord->id ) {
1.32 noro 3613: case 0:
1.43 noro 3614: switch ( ord->ord.simple ) {
3615: case 0:
3616: nd_dcomp = 1;
3617: nd_isrlex = 1;
3618: break;
3619: case 1:
3620: nd_dcomp = 1;
3621: nd_isrlex = 0;
3622: break;
3623: case 2:
3624: nd_dcomp = 0;
3625: nd_isrlex = 0;
1.45 noro 3626: ndl_compare_function = ndl_lex_compare;
1.58 noro 3627: break;
3628: case 11:
3629: /* XXX */
3630: nd_dcomp = 0;
3631: nd_isrlex = 1;
3632: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3633: break;
3634: default:
3635: error("nd_gr : unsupported order");
3636: }
1.32 noro 3637: break;
3638: case 1:
1.43 noro 3639: /* XXX */
3640: nd_dcomp = -1;
1.32 noro 3641: nd_isrlex = 0;
1.45 noro 3642: ndl_compare_function = ndl_block_compare;
1.34 noro 3643: break;
1.43 noro 3644: case 2:
3645: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3646: break;
3647: }
1.41 noro 3648: nd_ord = ord;
1.32 noro 3649: }
3650:
1.43 noro 3651: BlockMask nd_create_blockmask(struct order_spec *ord)
3652: {
3653: int n,i,j,s,l;
1.61 noro 3654: UINT *t;
1.43 noro 3655: BlockMask bm;
3656:
3657: if ( !ord->id )
3658: return 0;
3659: n = ord->ord.block.length;
3660: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3661: bm->n = n;
3662: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3663: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3664: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3665: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3666: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3667: l = bm->order_pair[i].length;
3668: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3669: }
3670: return bm;
1.57 noro 3671: }
3672:
3673: EPOS nd_create_epos(struct order_spec *ord)
3674: {
3675: int i,j,l,s,ord_l,ord_o;
3676: EPOS epos;
3677: struct order_pair *op;
3678:
3679: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3680: switch ( ord->id ) {
3681: case 0:
3682: if ( nd_isrlex ) {
3683: for ( i = 0; i < nd_nvar; i++ ) {
3684: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3685: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3686: }
3687: } else {
3688: for ( i = 0; i < nd_nvar; i++ ) {
3689: epos[i].i = nd_exporigin + i/nd_epw;
3690: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3691: }
3692: }
3693: break;
3694: case 1:
3695: /* block order */
3696: l = ord->ord.block.length;
3697: op = ord->ord.block.order_pair;
3698: for ( j = 0, s = 0; j < l; j++ ) {
3699: ord_o = op[j].order;
3700: ord_l = op[j].length;
3701: if ( !ord_o )
3702: for ( i = 0; i < ord_l; i++ ) {
3703: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3704: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3705: }
3706: else
3707: for ( i = 0; i < ord_l; i++ ) {
3708: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3709: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3710: }
3711: s += ord_l;
3712: }
3713: break;
3714: case 2:
3715: error("nd_create_epos : matrix order is not supported yet.");
3716: }
3717: return epos;
1.43 noro 3718: }
1.59 noro 3719:
3720: /* external interface */
3721:
3722: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3723: {
1.61 noro 3724: NODE t,in0,in;
1.59 noro 3725: ND nd,nf;
1.61 noro 3726: NDV ndv;
3727: VL vv,tv;
3728: int stat,nvar,max,e;
1.69 noro 3729: union oNDC dn;
1.59 noro 3730:
3731: pltovl(v,&vv);
1.74 noro 3732: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3733:
3734: /* get the degree bound */
3735: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3736: for ( tv = vv; tv; tv = NEXT(tv) ) {
3737: e = getdeg(tv->v,(P)BDY(t));
3738: max = MAX(e,max);
3739: }
3740: for ( tv = vv; tv; tv = NEXT(tv) ) {
3741: e = getdeg(tv->v,f);
3742: max = MAX(e,max);
3743: }
3744:
1.59 noro 3745: nd_init_ord(ord);
1.61 noro 3746: nd_setup_parameters(nvar,max);
3747:
3748: /* conversion to ndv */
3749: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3750: NEXTNODE(in0,in);
3751: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3752: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3753: }
3754: NEXTNODE(in0,in);
3755: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3756: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3757: NEXT(in) = 0;
3758:
3759: ndv_setup(m,0,in0);
1.59 noro 3760: nd_psn--;
3761: nd_scale=2;
3762: while ( 1 ) {
3763: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3764: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3765: if ( !stat ) {
3766: nd_psn++;
3767: nd_reconstruct(m,0,0);
3768: nd_psn--;
3769: } else
3770: break;
3771: }
1.61 noro 3772: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3773: }
3774:
3775: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3776: {
3777: NM m;
3778: UINT *t,*s;
3779: int i;
3780:
3781: for ( i = 0; i < n; i++ ) r[i] = 0;
3782: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3783: t = DL(m);
3784: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3785: r[i] = CM(m);
3786: }
3787: for ( i = 0; !r[i]; i++ );
3788: return i;
3789: }
3790:
1.74 noro 3791: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3792: {
3793: NMV m;
3794: UINT *t,*s;
3795: int i,j,len;
3796:
3797: for ( i = 0; i < n; i++ ) r[i] = 0;
3798: m = BDY(d);
3799: len = LEN(d);
3800: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3801: t = DL(m);
3802: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3803: r[i] = CM(m);
3804: }
3805: for ( i = 0; !r[i]; i++ );
3806: return i;
3807: }
3808:
1.63 noro 3809: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3810: {
3811: NM m;
3812: NMV mr;
3813: UINT *d,*t,*s;
3814: NDV p;
3815: int i,j,len;
3816:
3817: m = pair->mul;
3818: d = DL(m);
3819: p = nd_ps[pair->index];
3820: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3821: for ( i = 0; i < n; i++ ) r[i] = 0;
3822: len = LEN(p);
3823: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3824: ndl_add(d,DL(mr),t);
3825: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3826: r[i] = CM(mr);
3827: }
3828: for ( i = 0; !r[i]; i++ );
3829: return i;
3830: }
3831:
1.67 noro 3832: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3833: {
3834: NM m;
3835: NMV mr;
3836: UINT *d,*t,*s;
3837: NDV p;
1.67 noro 3838: unsigned char *ivc;
3839: unsigned short *ivs;
1.81 ! noro 3840: UINT *v,*ivi,*s0v;
1.67 noro 3841: int i,j,len,prev,diff,cdiff;
3842: IndArray r;
1.64 noro 3843:
3844: m = pair->mul;
3845: d = DL(m);
3846: p = nd_ps[pair->index];
3847: len = LEN(p);
3848: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3849: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3850: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3851: ndl_add(d,DL(mr),t);
3852: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3853: v[j] = i;
3854: }
1.81 ! noro 3855: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 3856: r->head = v[0];
3857: diff = 0;
3858: for ( i = 1; i < len; i++ ) {
3859: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3860: }
3861: if ( diff < 256 ) {
3862: r->width = 1;
3863: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3864: r->index.c = ivc;
3865: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3866: } else if ( diff < 65536 ) {
3867: r->width = 2;
3868: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3869: r->index.s = ivs;
3870: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3871: } else {
3872: r->width = 4;
3873: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3874: r->index.i = ivi;
3875: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3876: }
1.67 noro 3877: return r;
1.64 noro 3878: }
3879:
1.65 noro 3880:
1.76 noro 3881: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3882: {
1.67 noro 3883: int i,j,k,len,pos,prev;
1.66 noro 3884: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3885: IndArray ivect;
3886: unsigned char *ivc;
3887: unsigned short *ivs;
3888: unsigned int *ivi;
1.65 noro 3889: NDV redv;
1.67 noro 3890: NMV mr;
1.65 noro 3891: NODE rp;
1.76 noro 3892: int maxrs;
1.65 noro 3893:
1.76 noro 3894: maxrs = 0;
1.74 noro 3895: for ( i = 0; i < nred; i++ ) {
1.65 noro 3896: ivect = imat[i];
1.67 noro 3897: k = ivect->head; svect[k] %= m;
1.65 noro 3898: if ( c = svect[k] ) {
1.76 noro 3899: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3900: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3901: len = LEN(redv); mr = BDY(redv);
3902: svect[k] = 0; prev = k;
3903: switch ( ivect->width ) {
3904: case 1:
3905: ivc = ivect->index.c;
3906: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3907: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3908: prev = pos;
3909: DMA(c1,c,c2,up,lo);
3910: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3911: } else svect[pos] = lo;
3912: }
3913: break;
3914: case 2:
3915: ivs = ivect->index.s;
3916: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3917: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3918: prev = pos;
3919: DMA(c1,c,c2,up,lo);
3920: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3921: } else svect[pos] = lo;
3922: }
3923: break;
3924: case 4:
3925: ivi = ivect->index.i;
3926: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3927: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3928: prev = pos;
3929: DMA(c1,c,c2,up,lo);
3930: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3931: } else svect[pos] = lo;
3932: }
3933: break;
1.65 noro 3934: }
3935: }
3936: }
1.66 noro 3937: for ( i = 0; i < col; i++ )
3938: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3939: return maxrs;
1.65 noro 3940: }
3941:
1.76 noro 3942: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3943: {
3944: int i,j,k,len,pos,prev;
3945: UINT c,c1,c2,c3,up,lo,dmy;
3946: IndArray ivect;
3947: unsigned char *ivc;
3948: unsigned short *ivs;
3949: unsigned int *ivi;
3950: NDV redv;
3951: NMV mr;
3952: NODE rp;
1.76 noro 3953: int maxrs;
1.72 noro 3954:
1.76 noro 3955: maxrs = 0;
1.74 noro 3956: for ( i = 0; i < nred; i++ ) {
1.72 noro 3957: ivect = imat[i];
3958: k = ivect->head; svect[k] %= m;
3959: if ( c = svect[k] ) {
1.76 noro 3960: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3961: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3962: len = LEN(redv); mr = BDY(redv);
3963: svect[k] = 0; prev = k;
3964: switch ( ivect->width ) {
3965: case 1:
3966: ivc = ivect->index.c;
3967: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3968: pos = prev+ivc[j]; prev = pos;
3969: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3970: }
3971: break;
3972: case 2:
3973: ivs = ivect->index.s;
3974: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3975: pos = prev+ivs[j]; prev = pos;
3976: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3977: }
3978: break;
3979: case 4:
3980: ivi = ivect->index.i;
3981: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3982: pos = prev+ivi[j]; prev = pos;
3983: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3984: }
3985: break;
3986: }
3987: }
3988: }
1.76 noro 3989: return maxrs;
1.72 noro 3990: }
3991:
1.65 noro 3992: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3993: {
3994: int j,k,len;
3995: UINT *p;
3996: UINT c;
3997: NDV r;
3998: NMV mr0,mr;
3999:
4000: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4001: if ( !len ) return 0;
4002: else {
1.74 noro 4003: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4004: ndv_alloc += nmv_adv*len;
1.65 noro 4005: mr = mr0;
4006: p = s0vect;
4007: for ( j = k = 0; j < col; j++, p += nd_wpd )
4008: if ( !rhead[j] ) {
4009: if ( c = vect[k++] ) {
4010: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4011: }
4012: }
4013: MKNDV(nd_nvar,mr0,len,r);
4014: return r;
4015: }
4016: }
4017:
1.74 noro 4018: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4019: {
4020: ND_pairs t;
4021: NODE sp0,sp;
4022: int stat;
4023: ND spol;
4024:
4025: for ( t = l; t; t = NEXT(t) ) {
4026: stat = nd_sp(m,0,t,&spol);
4027: if ( !stat ) return 0;
4028: if ( spol ) {
4029: add_pbucket_symbolic(bucket,spol);
4030: }
4031: }
1.68 noro 4032: return 1;
1.65 noro 4033: }
4034:
4035: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4036: {
4037: NODE rp0,rp;
4038: NM mul,head,s0,s;
1.76 noro 4039: int index,col,i,sugar;
1.65 noro 4040: RHist h;
4041: UINT *s0v,*p;
4042: NM_ind_pair pair;
4043: ND red;
4044:
4045: s0 = 0; rp0 = 0; col = 0;
4046: while ( 1 ) {
4047: head = remove_head_pbucket_symbolic(bucket);
4048: if ( !head ) break;
4049: if ( !s0 ) s0 = head;
4050: else NEXT(s) = head;
4051: s = head;
4052: index = ndl_find_reducer(DL(head));
4053: if ( index >= 0 ) {
4054: h = nd_psh[index];
4055: NEWNM(mul);
4056: ndl_sub(DL(head),DL(h),DL(mul));
4057: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4058: sugar = TD(DL(mul))+SG(nd_ps[index]);
4059: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4060: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4061: add_pbucket_symbolic(bucket,nd_remove_head(red));
4062: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4063: }
4064: col++;
4065: }
1.72 noro 4066: if ( rp0 ) NEXT(rp) = 0;
4067: NEXT(s) = 0;
1.65 noro 4068: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4069: for ( i = 0, p = s0v, s = s0; i < col;
4070: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4071: *s0vect = s0v;
4072: *r = rp0;
4073: return col;
4074: }
4075:
1.69 noro 4076: NODE nd_f4(int m)
4077: {
4078: int i,nh,stat,index;
4079: NODE r,g;
4080: ND_pairs d,l,t;
4081: ND spol,red;
4082: NDV nf,redv;
4083: NM s0,s;
1.74 noro 4084: NODE rp0,sp0,srp0,nflist;
1.69 noro 4085: int nsp,nred,col,rank,len,k,j,a;
4086: UINT c;
1.74 noro 4087: UINT **spmat;
1.69 noro 4088: UINT *s0vect,*svect,*p,*v;
4089: int *colstat;
4090: IndArray *imat;
4091: int *rhead;
4092: int spcol,sprow;
4093: int sugar;
4094: PGeoBucket bucket;
4095: struct oEGT eg0,eg1,eg_f4;
4096:
4097: if ( !m )
4098: error("nd_f4 : not implemented");
1.74 noro 4099: ndv_alloc = 0;
1.69 noro 4100: g = 0; d = 0;
4101: for ( i = 0; i < nd_psn; i++ ) {
4102: d = update_pairs(d,g,i);
4103: g = update_base(g,i);
4104: }
4105: while ( d ) {
4106: get_eg(&eg0);
4107: l = nd_minsugarp(d,&d);
4108: sugar = SG(l);
4109: bucket = create_pbucket();
1.74 noro 4110: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4111: if ( !stat ) {
4112: for ( t = l; NEXT(t); t = NEXT(t) );
4113: NEXT(t) = d; d = l;
4114: d = nd_reconstruct(m,0,d);
4115: continue;
4116: }
4117: if ( !sp0 ) continue;
4118: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4119: if ( !col ) {
4120: for ( t = l; NEXT(t); t = NEXT(t) );
4121: NEXT(t) = d; d = l;
4122: d = nd_reconstruct(m,0,d);
4123: continue;
4124: }
4125: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4126: if ( DP_Print )
4127: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4128: sugar,eg_f4.exectime+eg_f4.gctime);
4129: if ( 1 )
4130: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4131: else
4132: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4133: /* adding new bases */
1.74 noro 4134: for ( r = nflist; r; r = NEXT(r) ) {
4135: nf = (NDV)BDY(r);
1.69 noro 4136: ndv_removecont(m,nf);
1.77 noro 4137: nh = ndv_newps(m,nf,0);
1.69 noro 4138: d = update_pairs(d,g,nh);
4139: g = update_base(g,nh);
4140: }
4141: }
4142: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 4143: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 4144: return g;
4145: }
1.74 noro 4146:
4147: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4148: {
1.67 noro 4149: IndArray *imat;
1.74 noro 4150: int nsp,nred,spcol,sprow,a;
1.65 noro 4151: int *rhead;
1.74 noro 4152: int i,j,k,l,rank;
4153: NODE rp,r0,r;
4154: ND_pairs sp;
4155: ND spol;
4156: int **spmat;
4157: UINT *svect,*v;
4158: int *colstat;
1.80 noro 4159: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
1.74 noro 4160: NM_ind_pair *rvect;
1.76 noro 4161: int maxrs;
4162: int *spsugar;
1.63 noro 4163:
1.74 noro 4164: get_eg(&eg0);
4165: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4166: nred = length(rp0); spcol = col-nred;
4167: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4168: rhead = (int *)ALLOCA(col*sizeof(int));
4169: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4170:
1.74 noro 4171: /* construction of index arrays */
4172: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4173: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4174: rvect[i] = (NM_ind_pair)BDY(rp);
4175: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4176: rhead[imat[i]->head] = 1;
4177: }
4178:
4179: /* elimination (1st step) */
4180: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4181: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4182: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4183: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4184: nd_sp(m,0,sp,&spol);
1.75 noro 4185: if ( !spol ) continue;
1.74 noro 4186: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4187: if ( m == -1 )
4188: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4189: else
4190: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4191: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4192: if ( i < col ) {
4193: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4194: for ( j = k = 0; j < col; j++ )
4195: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4196: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4197: sprow++;
4198: }
1.76 noro 4199: nd_free(spol);
1.74 noro 4200: }
1.80 noro 4201: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4202: if ( DP_Print ) {
4203: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4204: fflush(asir_out);
4205: }
1.74 noro 4206: /* free index arrays */
4207: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4208:
4209: /* elimination (2nd step) */
4210: colstat = (int *)ALLOCA(spcol*sizeof(int));
4211: if ( m == -1 )
1.76 noro 4212: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4213: else
1.76 noro 4214: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4215: r0 = 0;
4216: for ( i = 0; i < rank; i++ ) {
4217: NEXTNODE(r0,r); BDY(r) =
4218: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4219: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4220: GC_free(spmat[i]);
4221: }
4222: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4223: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4224: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4225: if ( DP_Print ) {
1.80 noro 4226: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4227: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4228: nsp,nred,sprow,spcol,rank);
4229: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4230: }
1.74 noro 4231: return r0;
4232: }
4233:
4234: FILE *nd_write,*nd_read;
4235:
4236: void nd_send_int(int a) {
4237: write_int(nd_write,&a);
4238: }
4239:
4240: void nd_send_intarray(int *p,int len) {
4241: write_intarray(nd_write,p,len);
4242: }
4243:
4244: int nd_recv_int() {
4245: int a;
4246:
4247: read_int(nd_read,&a);
4248: return a;
4249: }
4250:
4251: void nd_recv_intarray(int *p,int len) {
4252: read_intarray(nd_read,p,len);
4253: }
4254:
4255: void nd_send_ndv(NDV p) {
4256: int len,i;
4257: NMV m;
4258:
4259: if ( !p ) nd_send_int(0);
4260: else {
4261: len = LEN(p);
4262: nd_send_int(len);
4263: m = BDY(p);
4264: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4265: nd_send_int(CM(m));
4266: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4267: }
1.74 noro 4268: }
4269: }
4270:
4271: void nd_send_nd(ND p) {
4272: int len,i;
4273: NM m;
4274:
4275: if ( !p ) nd_send_int(0);
4276: else {
4277: len = LEN(p);
4278: nd_send_int(len);
4279: m = BDY(p);
4280: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4281: nd_send_int(CM(m));
4282: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4283: }
1.74 noro 4284: }
4285: }
1.65 noro 4286:
1.74 noro 4287: NDV nd_recv_ndv()
4288: {
4289: int len,i;
4290: NMV m,m0;
4291: NDV r;
1.65 noro 4292:
1.74 noro 4293: len = nd_recv_int();
4294: if ( !len ) return 0;
4295: else {
4296: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4297: ndv_alloc += len*nmv_adv;
4298: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4299: CM(m) = nd_recv_int();
4300: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4301: }
1.74 noro 4302: MKNDV(nd_nvar,m0,len,r);
4303: return r;
4304: }
4305: }
1.65 noro 4306:
1.74 noro 4307: int ox_exec_f4_red(Q proc)
4308: {
4309: Obj obj;
4310: STRING fname;
4311: NODE arg;
4312: int s;
4313: extern int ox_need_conv,ox_file_io;
4314:
4315: MKSTR(fname,"nd_exec_f4_red");
4316: arg = mknode(2,proc,fname);
4317: Pox_cmo_rpc(arg,&obj);
4318: s = get_ox_server_id(QTOS(proc));
4319: nd_write = iofp[s].out;
4320: nd_read = iofp[s].in;
4321: ox_need_conv = ox_file_io = 0;
4322: return s;
4323: }
4324:
4325: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4326: {
4327: int nsp,nred;
4328: int i,rank,s;
4329: NODE rp,r0,r;
4330: ND_pairs sp;
4331: NM_ind_pair pair;
4332: NMV nmv;
4333: NM nm;
4334: NDV nf;
4335: Obj proc,dmy;
4336:
4337: ox_launch_main(0,0,&proc);
4338: s = ox_exec_f4_red((Q)proc);
4339:
4340: nd_send_int(m);
4341: nd_send_int(nd_nvar);
4342: nd_send_int(nd_bpe);
4343: nd_send_int(nd_wpd);
4344: nd_send_int(nmv_adv);
4345:
4346: saveobj(nd_write,dp_current_spec.obj); fflush(nd_write);
4347:
4348: nd_send_int(nd_psn);
4349: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4350:
4351: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4352: nd_send_int(nsp);
4353: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4354: nd_send_int(sp->i1); nd_send_int(sp->i2);
4355: }
4356:
4357: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4358:
4359: nred = length(rp0); nd_send_int(nred);
4360: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4361: pair = (NM_ind_pair)BDY(rp);
4362: nd_send_int(pair->index);
4363: nd_send_intarray(pair->mul->dl,nd_wpd);
4364: }
4365: fflush(nd_write);
4366: rank = nd_recv_int();
4367: fprintf(asir_out,"rank=%d\n",rank);
4368: r0 = 0;
4369: for ( i = 0; i < rank; i++ ) {
4370: nf = nd_recv_ndv();
4371: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4372: }
4373: Pox_shutdown(mknode(1,proc),&dmy);
4374: return r0;
4375: }
4376:
4377: /* server side */
4378:
4379: void nd_exec_f4_red_dist()
4380: {
4381: int m,i,nsp,col,s0size,nred,spcol,j,k;
4382: NM_ind_pair *rp0;
4383: NDV nf;
4384: UINT *s0vect;
4385: IndArray *imat;
4386: int *rhead;
4387: int **spmat;
4388: UINT *svect,*v;
4389: ND_pairs *sp0;
4390: int *colstat;
4391: int a,sprow,rank;
4392: struct order_spec ord;
4393: Obj ordspec;
4394: ND spol;
1.76 noro 4395: int maxrs;
4396: int *spsugar;
1.74 noro 4397:
4398: nd_read = iofp[0].in;
4399: nd_write = iofp[0].out;
4400: m = nd_recv_int();
4401: nd_nvar = nd_recv_int();
4402: nd_bpe = nd_recv_int();
4403: nd_wpd = nd_recv_int();
4404: nmv_adv = nd_recv_int();
4405:
4406: loadobj(nd_read,&ordspec);
4407: create_order_spec(ordspec,&ord);
4408: nd_init_ord(&ord);
4409: nd_setup_parameters(nd_nvar,0);
4410:
4411: nd_psn = nd_recv_int();
4412: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4413: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4414: for ( i = 0; i < nd_psn; i++ ) {
4415: nd_ps[i] = nd_recv_ndv();
4416: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4417: }
4418:
4419: nsp = nd_recv_int();
4420: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4421: for ( i = 0; i < nsp; i++ ) {
4422: NEWND_pairs(sp0[i]);
4423: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4424: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4425: }
4426:
4427: col = nd_recv_int();
4428: s0size = col*nd_wpd;
4429: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4430: nd_recv_intarray(s0vect,s0size);
4431:
4432: nred = nd_recv_int();
4433: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4434: for ( i = 0; i < nred; i++ ) {
4435: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4436: rp0[i]->index = nd_recv_int();
4437: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4438: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4439: }
4440:
4441: spcol = col-nred;
4442: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4443: rhead = (int *)MALLOC(col*sizeof(int));
4444: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4445:
4446: /* construction of index arrays */
4447: for ( i = 0; i < nred; i++ ) {
4448: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4449: rhead[imat[i]->head] = 1;
4450: }
4451:
4452: /* elimination (1st step) */
4453: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4454: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4455: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4456: for ( a = sprow = 0; a < nsp; a++ ) {
4457: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4458: if ( !spol ) continue;
1.74 noro 4459: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4460: if ( m == -1 )
4461: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4462: else
4463: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4464: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4465: if ( i < col ) {
4466: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4467: for ( j = k = 0; j < col; j++ )
4468: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4469: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4470: sprow++;
4471: }
1.76 noro 4472: nd_free(spol);
1.74 noro 4473: }
4474: /* elimination (2nd step) */
4475: colstat = (int *)ALLOCA(spcol*sizeof(int));
4476: if ( m == -1 )
1.76 noro 4477: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4478: else
1.76 noro 4479: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4480: nd_send_int(rank);
4481: for ( i = 0; i < rank; i++ ) {
4482: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4483: nd_send_ndv(nf);
1.63 noro 4484: }
1.74 noro 4485: fflush(nd_write);
1.76 noro 4486: }
4487:
4488: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4489: {
4490: int i,j,k,l,inv,a,rank,s;
4491: unsigned int *t,*pivot,*pk;
4492: unsigned int **mat;
4493:
4494: mat = (unsigned int **)mat0;
4495: for ( rank = 0, j = 0; j < col; j++ ) {
4496: for ( i = rank; i < row; i++ )
4497: mat[i][j] %= md;
4498: for ( i = rank; i < row; i++ )
4499: if ( mat[i][j] )
4500: break;
4501: if ( i == row ) {
4502: colstat[j] = 0;
4503: continue;
4504: } else
4505: colstat[j] = 1;
4506: if ( i != rank ) {
4507: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4508: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4509: }
4510: pivot = mat[rank];
4511: s = sugar[rank];
4512: inv = invm(pivot[j],md);
4513: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4514: if ( *pk ) {
4515: if ( *pk >= (unsigned int)md )
4516: *pk %= md;
4517: DMAR(*pk,inv,0,md,*pk)
4518: }
4519: for ( i = rank+1; i < row; i++ ) {
4520: t = mat[i];
4521: if ( a = t[j] ) {
4522: sugar[i] = MAX(sugar[i],s);
4523: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4524: }
4525: }
4526: rank++;
4527: }
4528: for ( j = col-1, l = rank-1; j >= 0; j-- )
4529: if ( colstat[j] ) {
4530: pivot = mat[l];
4531: s = sugar[l];
4532: for ( i = 0; i < l; i++ ) {
4533: t = mat[i];
4534: t[j] %= md;
4535: if ( a = t[j] ) {
4536: sugar[i] = MAX(sugar[i],s);
4537: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4538: }
4539: }
4540: l--;
4541: }
4542: for ( j = 0, l = 0; l < rank; j++ )
4543: if ( colstat[j] ) {
4544: t = mat[l];
4545: for ( k = j; k < col; k++ )
4546: if ( t[k] >= (unsigned int)md )
4547: t[k] %= md;
4548: l++;
4549: }
4550: return rank;
4551: }
4552:
4553: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4554: {
4555: int i,j,k,l,inv,a,rank,s;
4556: unsigned int *t,*pivot,*pk;
4557: unsigned int **mat;
4558:
4559: mat = (unsigned int **)mat0;
4560: for ( rank = 0, j = 0; j < col; j++ ) {
4561: for ( i = rank; i < row; i++ )
4562: if ( mat[i][j] )
4563: break;
4564: if ( i == row ) {
4565: colstat[j] = 0;
4566: continue;
4567: } else
4568: colstat[j] = 1;
4569: if ( i != rank ) {
4570: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4571: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4572: }
4573: pivot = mat[rank];
4574: s = sugar[rank];
4575: inv = _invsf(pivot[j]);
4576: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4577: if ( *pk )
4578: *pk = _mulsf(*pk,inv);
4579: for ( i = rank+1; i < row; i++ ) {
4580: t = mat[i];
4581: if ( a = t[j] ) {
4582: sugar[i] = MAX(sugar[i],s);
4583: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4584: }
4585: }
4586: rank++;
4587: }
4588: for ( j = col-1, l = rank-1; j >= 0; j-- )
4589: if ( colstat[j] ) {
4590: pivot = mat[l];
4591: s = sugar[l];
4592: for ( i = 0; i < l; i++ ) {
4593: t = mat[i];
4594: if ( a = t[j] ) {
4595: sugar[i] = MAX(sugar[i],s);
4596: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4597: }
4598: }
4599: l--;
4600: }
4601: return rank;
1.77 noro 4602: }
4603:
4604: int ndv_ishomo(NDV p)
4605: {
4606: NMV m;
4607: int len,h;
4608:
4609: if ( !p ) return 1;
4610: len = LEN(p);
4611: m = BDY(p);
4612: h = TD(DL(m));
4613: NMV_ADV(m);
4614: for ( len--; len; len--, NMV_ADV(m) )
4615: if ( TD(DL(m)) != h ) return 0;
4616: return 1;
4617: }
4618:
4619: void ndv_save(NDV p,int index)
4620: {
4621: FILE *s;
4622: char name[BUFSIZ];
4623: short id;
4624: int nv,sugar,len,n,i,td,e,j;
4625: NMV m;
4626: unsigned int *dl;
4627:
4628: sprintf(name,"%s/%d",Demand,index);
4629: s = fopen(name,"w");
4630: savevl(s,0);
4631: if ( !p ) {
4632: saveobj(s,0);
4633: return;
4634: }
4635: id = O_DP;
4636: nv = NV(p);
4637: sugar = SG(p);
4638: len = LEN(p);
4639: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
4640: write_int(s,&len);
4641:
4642: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4643: saveobj(s,(Obj)CQ(m));
4644: dl = DL(m);
4645: td = TD(dl);
4646: write_int(s,&td);
4647: for ( j = 0; j < nv; j++ ) {
4648: e = GET_EXP(dl,j);
4649: write_int(s,&e);
4650: }
4651: }
4652: fclose(s);
4653: }
4654:
4655: NDV ndv_load(int index)
4656: {
4657: FILE *s;
4658: char name[BUFSIZ];
4659: short id;
4660: int nv,sugar,len,n,i,td,e,j;
4661: NDV d;
4662: NMV m0,m;
4663: unsigned int *dl;
4664: Obj obj;
4665:
4666: sprintf(name,"%s/%d",Demand,index);
4667: s = fopen(name,"r");
4668: if ( !s ) return 0;
4669:
4670: skipvl(s);
4671: read_short(s,&id);
4672: if ( !id ) return 0;
4673: read_int(s,&nv);
4674: read_int(s,&sugar);
4675: read_int(s,&len);
4676:
4677: m0 = m = MALLOC(len*nmv_adv);
4678: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4679: loadobj(s,&obj); CQ(m) = (Q)obj;
4680: dl = DL(m);
4681: ndl_zero(dl);
4682: read_int(s,&td); TD(dl) = td;
4683: for ( j = 0; j < nv; j++ ) {
4684: read_int(s,&e);
4685: PUT_EXP(dl,j,e);
4686: }
4687: if ( nd_blockmask ) ndl_weight_mask(dl);
4688: }
4689: fclose(s);
4690: MKNDV(nv,m0,len,d);
4691: SG(d) = sugar;
4692: return d;
1.59 noro 4693: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>