Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.77
1.77 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.76 2003/09/28 09:18:57 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:
1.41 noro 787: for ( i = 0; i < nd_wpd; i++ )
1.34 noro 788: if ( *d1++ != *d2++ )
1.1 noro 789: return 0;
790: return 1;
791: }
792:
1.61 noro 793: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 794: {
795: int i;
796:
797: switch ( nd_wpd ) {
1.41 noro 798: case 2:
1.34 noro 799: TD(d2) = TD(d1);
800: d2[1] = d1[1];
1.6 noro 801: break;
1.41 noro 802: case 3:
1.34 noro 803: TD(d2) = TD(d1);
1.6 noro 804: d2[1] = d1[1];
1.34 noro 805: d2[2] = d1[2];
1.6 noro 806: break;
807: default:
1.41 noro 808: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 809: d2[i] = d1[i];
810: break;
811: }
812: }
813:
1.61 noro 814: INLINE void ndl_zero(UINT *d)
815: {
816: int i;
817: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
818: }
819:
820: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 821: {
822: int i;
823:
1.43 noro 824: #if 1
1.6 noro 825: switch ( nd_wpd ) {
1.41 noro 826: case 2:
827: TD(d) = TD(d1)+TD(d2);
1.34 noro 828: d[1] = d1[1]+d2[1];
1.6 noro 829: break;
1.41 noro 830: case 3:
831: TD(d) = TD(d1)+TD(d2);
1.6 noro 832: d[1] = d1[1]+d2[1];
1.34 noro 833: d[2] = d1[2]+d2[2];
1.6 noro 834: break;
835: default:
1.43 noro 836: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 837: break;
838: }
1.43 noro 839: #else
840: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
841: #endif
1.6 noro 842: }
843:
1.55 noro 844: /* d1 += d2 */
1.61 noro 845: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 846: {
847: int i;
848:
849: #if 1
850: switch ( nd_wpd ) {
851: case 2:
852: TD(d1) += TD(d2);
853: d1[1] += d2[1];
854: break;
855: case 3:
856: TD(d1) += TD(d2);
857: d1[1] += d2[1];
858: d1[2] += d2[2];
859: break;
860: default:
861: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
862: break;
863: }
864: #else
865: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
866: #endif
867: }
868:
1.61 noro 869: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 870: {
871: int i;
872:
1.43 noro 873: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 874: }
875:
1.61 noro 876: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 877: {
1.61 noro 878: UINT t1,t2,u,u1,u2;
1.1 noro 879: int i,j;
880:
1.65 noro 881: #if USE_UNROLL
1.1 noro 882: switch ( nd_bpe ) {
1.62 noro 883: case 3:
884: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
885: u1 = d1[i]; u2 = d2[i];
886: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
887: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
888: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
889: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
890: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
891: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
892: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
893: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
894: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
895: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
896: }
897: return 1;
898: break;
1.1 noro 899: case 4:
1.41 noro 900: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 901: u1 = d1[i]; u2 = d2[i];
902: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 903: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
904: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
905: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
906: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
907: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
908: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
909: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 910: }
911: return 1;
912: break;
913: case 6:
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&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 917: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
918: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
919: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
920: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 921: }
922: return 1;
923: break;
924: case 8:
1.41 noro 925: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 926: u1 = d1[i]; u2 = d2[i];
927: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 928: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
929: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
930: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 931: }
932: return 1;
933: break;
934: case 16:
1.41 noro 935: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 936: u1 = d1[i]; u2 = d2[i];
937: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 938: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 939: }
940: return 1;
941: break;
942: case 32:
1.41 noro 943: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 944: if ( d1[i] && d2[i] ) return 0;
945: return 1;
946: break;
947: default:
1.41 noro 948: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 949: u1 = d1[i]; u2 = d2[i];
950: for ( j = 0; j < nd_epw; j++ ) {
951: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
952: u1 >>= nd_bpe; u2 >>= nd_bpe;
953: }
954: }
955: return 1;
956: break;
957: }
1.65 noro 958: #else
959: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
960: u1 = d1[i]; u2 = d2[i];
961: for ( j = 0; j < nd_epw; j++ ) {
962: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
963: u1 >>= nd_bpe; u2 >>= nd_bpe;
964: }
965: }
966: return 1;
967: #endif
1.1 noro 968: }
969:
1.61 noro 970: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 971: {
1.61 noro 972: UINT u2;
973: UINT *d1;
1.5 noro 974: int i,j,ind,k;
1.1 noro 975:
1.5 noro 976: d1 = nd_bound[index];
977: ind = 0;
1.65 noro 978: #if USE_UNROLL
1.5 noro 979: switch ( nd_bpe ) {
1.62 noro 980: case 3:
981: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
982: u2 = d2[i];
983: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
984: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
985: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
986: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
987: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
988: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
989: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
990: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
991: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
992: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
993: }
994: return 0;
995: break;
1.5 noro 996: case 4:
1.41 noro 997: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 998: u2 = d2[i];
999: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
1000: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
1001: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
1002: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
1003: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
1004: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
1005: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1006: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1007: }
1008: return 0;
1009: break;
1010: case 6:
1.41 noro 1011: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1012: u2 = d2[i];
1013: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1014: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1015: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1016: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1017: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1018: }
1019: return 0;
1020: break;
1021: case 8:
1.41 noro 1022: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1023: u2 = d2[i];
1024: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1025: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1026: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1027: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1028: }
1029: return 0;
1030: break;
1031: case 16:
1.41 noro 1032: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1033: u2 = d2[i];
1034: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1035: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1036: }
1037: return 0;
1038: break;
1039: case 32:
1.41 noro 1040: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1041: if ( d1[i]+d2[i]<d1[i] ) return 1;
1042: return 0;
1043: break;
1044: default:
1.41 noro 1045: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1046: u2 = d2[i];
1047: k = (nd_epw-1)*nd_bpe;
1048: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1049: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1050: }
1051: return 0;
1052: break;
1053: }
1.65 noro 1054: #else
1055: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1056: u2 = d2[i];
1057: k = (nd_epw-1)*nd_bpe;
1058: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1059: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1060: }
1061: return 0;
1062: #endif
1.1 noro 1063: }
1064:
1.61 noro 1065: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1066: {
1067: int i;
1068: int r;
1069:
1.34 noro 1070: r = 0;
1.41 noro 1071: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1072: r = ((r<<16)+d[i])%REDTAB_LEN;
1073: return r;
1074: }
1075:
1.63 noro 1076: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1077: {
1.13 noro 1078: RHist r;
1.6 noro 1079: int d,k,i;
1.1 noro 1080:
1.63 noro 1081: d = ndl_hash_value(dg);
1.13 noro 1082: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1083: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1084: if ( k > 0 ) nd_notfirst++;
1085: nd_found++;
1.13 noro 1086: return r->index;
1.1 noro 1087: }
1088: }
1.13 noro 1089: if ( Reverse )
1090: for ( i = nd_psn-1; i >= 0; i-- ) {
1091: r = nd_psh[i];
1.34 noro 1092: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1093: nd_create++;
1.34 noro 1094: nd_append_red(dg,i);
1.13 noro 1095: return i;
1096: }
1097: }
1098: else
1099: for ( i = 0; i < nd_psn; i++ ) {
1100: r = nd_psh[i];
1.34 noro 1101: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1102: nd_create++;
1.34 noro 1103: nd_append_red(dg,i);
1.13 noro 1104: return i;
1105: }
1.1 noro 1106: }
1.6 noro 1107: return -1;
1.1 noro 1108: }
1109:
1.63 noro 1110: ND nd_merge(ND p1,ND p2)
1111: {
1112: int n,c;
1113: int t,can,td1,td2;
1114: ND r;
1115: NM m1,m2,mr0,mr,s;
1116:
1117: if ( !p1 ) return p2;
1118: else if ( !p2 ) return p1;
1119: else {
1120: can = 0;
1121: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1122: c = DL_COMPARE(DL(m1),DL(m2));
1123: switch ( c ) {
1124: case 0:
1125: s = m1; m1 = NEXT(m1);
1126: can++; NEXTNM2(mr0,mr,s);
1127: s = m2; m2 = NEXT(m2); FREENM(s);
1128: break;
1129: case 1:
1130: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1131: break;
1132: case -1:
1133: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1134: break;
1135: }
1136: }
1137: if ( !mr0 )
1138: if ( m1 ) mr0 = m1;
1139: else if ( m2 ) mr0 = m2;
1140: else return 0;
1141: else if ( m1 ) NEXT(mr) = m1;
1142: else if ( m2 ) NEXT(mr) = m2;
1143: else NEXT(mr) = 0;
1144: BDY(p1) = mr0;
1145: SG(p1) = MAX(SG(p1),SG(p2));
1146: LEN(p1) = LEN(p1)+LEN(p2)-can;
1147: FREEND(p2);
1148: return p1;
1149: }
1150: }
1151:
1.31 noro 1152: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1153: {
1154: int n,c;
1.34 noro 1155: int t,can,td1,td2;
1.1 noro 1156: ND r;
1157: NM m1,m2,mr0,mr,s;
1158:
1.34 noro 1159: if ( !p1 ) return p2;
1160: else if ( !p2 ) return p1;
1.71 noro 1161: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 1162: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1163: else {
1.30 noro 1164: can = 0;
1.1 noro 1165: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1166: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1167: switch ( c ) {
1168: case 0:
1.19 noro 1169: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1170: if ( t < 0 ) t += mod;
1.1 noro 1171: s = m1; m1 = NEXT(m1);
1172: if ( t ) {
1.34 noro 1173: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1174: } else {
1.34 noro 1175: can += 2; FREENM(s);
1.1 noro 1176: }
1177: s = m2; m2 = NEXT(m2); FREENM(s);
1178: break;
1179: case 1:
1180: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1181: break;
1182: case -1:
1183: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1184: break;
1185: }
1186: }
1187: if ( !mr0 )
1.34 noro 1188: if ( m1 ) mr0 = m1;
1189: else if ( m2 ) mr0 = m2;
1190: else return 0;
1191: else if ( m1 ) NEXT(mr) = m1;
1192: else if ( m2 ) NEXT(mr) = m2;
1193: else NEXT(mr) = 0;
1.1 noro 1194: BDY(p1) = mr0;
1.14 noro 1195: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1196: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1197: FREEND(p2);
1198: return p1;
1199: }
1200: }
1201:
1.31 noro 1202: ND nd_add_q(ND p1,ND p2)
1.17 noro 1203: {
1.30 noro 1204: int n,c,can;
1.17 noro 1205: ND r;
1206: NM m1,m2,mr0,mr,s;
1207: Q t;
1208:
1.34 noro 1209: if ( !p1 ) return p2;
1210: else if ( !p2 ) return p1;
1.31 noro 1211: else {
1.30 noro 1212: can = 0;
1.17 noro 1213: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1214: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1215: switch ( c ) {
1216: case 0:
1217: addq(CQ(m1),CQ(m2),&t);
1218: s = m1; m1 = NEXT(m1);
1219: if ( t ) {
1.34 noro 1220: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1221: } else {
1.34 noro 1222: can += 2; FREENM(s);
1.17 noro 1223: }
1224: s = m2; m2 = NEXT(m2); FREENM(s);
1225: break;
1226: case 1:
1227: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1228: break;
1229: case -1:
1230: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1231: break;
1232: }
1233: }
1234: if ( !mr0 )
1.34 noro 1235: if ( m1 ) mr0 = m1;
1236: else if ( m2 ) mr0 = m2;
1237: else return 0;
1238: else if ( m1 ) NEXT(mr) = m1;
1239: else if ( m2 ) NEXT(mr) = m2;
1240: else NEXT(mr) = 0;
1.17 noro 1241: BDY(p1) = mr0;
1242: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1243: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1244: FREEND(p2);
1245: return p1;
1246: }
1247: }
1248:
1.71 noro 1249: ND nd_add_sf(ND p1,ND p2)
1250: {
1251: int n,c,can;
1252: ND r;
1253: NM m1,m2,mr0,mr,s;
1254: int t;
1255:
1256: if ( !p1 ) return p2;
1257: else if ( !p2 ) return p1;
1258: else {
1259: can = 0;
1260: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1261: c = DL_COMPARE(DL(m1),DL(m2));
1262: switch ( c ) {
1263: case 0:
1264: t = _addsf(CM(m1),CM(m2));
1265: s = m1; m1 = NEXT(m1);
1266: if ( t ) {
1267: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1268: } else {
1269: can += 2; FREENM(s);
1270: }
1271: s = m2; m2 = NEXT(m2); FREENM(s);
1272: break;
1273: case 1:
1274: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1275: break;
1276: case -1:
1277: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1278: break;
1279: }
1280: }
1281: if ( !mr0 )
1282: if ( m1 ) mr0 = m1;
1283: else if ( m2 ) mr0 = m2;
1284: else return 0;
1285: else if ( m1 ) NEXT(mr) = m1;
1286: else if ( m2 ) NEXT(mr) = m2;
1287: else NEXT(mr) = 0;
1288: BDY(p1) = mr0;
1289: SG(p1) = MAX(SG(p1),SG(p2));
1290: LEN(p1) = LEN(p1)+LEN(p2)-can;
1291: FREEND(p2);
1292: return p1;
1293: }
1294: }
1295:
1.1 noro 1296: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1297: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1298: {
1.11 noro 1299: ND d;
1.1 noro 1300: NM m,mrd,tail;
1.7 noro 1301: NM mul;
1.10 noro 1302: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1303: int c,c1,c2,dummy;
1.17 noro 1304: RHist h;
1.11 noro 1305: NDV p,red;
1.69 noro 1306: Q cg,cred,gcd,tq,qq;
1.21 noro 1307: double hmag;
1.1 noro 1308:
1.69 noro 1309: if ( dn ) {
1310: if ( mod )
1311: dn->m = 1;
1312: else
1313: dn->z = ONE;
1314: }
1.1 noro 1315: if ( !g ) {
1316: *rp = 0;
1317: return 1;
1318: }
1.34 noro 1319: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1320:
1.14 noro 1321: sugar0 = sugar = SG(g);
1.1 noro 1322: n = NV(g);
1.61 noro 1323: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1324: for ( d = 0; g; ) {
1.63 noro 1325: index = ndl_find_reducer(HDL(g));
1.6 noro 1326: if ( index >= 0 ) {
1.17 noro 1327: h = nd_psh[index];
1328: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1329: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1330: nd_free(g); nd_free(d);
1331: return 0;
1332: }
1.77 ! noro 1333: if ( nd_demand )
! 1334: p = ndv_load(index);
! 1335: else
! 1336: p = ps[index];
1.71 noro 1337: if ( mod == -1 )
1338: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1339: else if ( mod ) {
1.19 noro 1340: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1341: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1342: } else {
1.17 noro 1343: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1344: chsgnq(cg,&CQ(mul));
1.20 noro 1345: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1346: if ( dn ) {
1347: mulq(dn->z,cred,&tq); dn->z = tq;
1348: }
1.16 noro 1349: }
1.55 noro 1350: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1351: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1352: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1353: tq = HCQ(g);
1.21 noro 1354: nd_removecont2(d,g);
1.69 noro 1355: if ( dn ) {
1356: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1357: }
1.21 noro 1358: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1359: }
1.1 noro 1360: } else if ( !full ) {
1361: *rp = g;
1362: return 1;
1363: } else {
1364: m = BDY(g);
1365: if ( NEXT(m) ) {
1.34 noro 1366: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1367: } else {
1368: FREEND(g); g = 0;
1369: }
1370: if ( d ) {
1.34 noro 1371: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1372: } else {
1.34 noro 1373: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1374: }
1375: }
1376: }
1.34 noro 1377: if ( d ) SG(d) = sugar;
1.1 noro 1378: *rp = d;
1379: return 1;
1380: }
1.28 noro 1381:
1.53 noro 1382: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1383: {
1384: int hindex,index;
1385: NDV p;
1386: ND u,d,red;
1387: NODE l;
1.31 noro 1388: NM mul,m,mrd,tail;
1.25 noro 1389: int sugar,psugar,n,h_reducible;
1390: PGeoBucket bucket;
1391: int c,c1,c2;
1.26 noro 1392: Q cg,cred,gcd,zzz;
1.25 noro 1393: RHist h;
1.28 noro 1394: double hmag,gmag;
1.77 ! noro 1395: int count = 0;
! 1396: int hcount = 0;
1.25 noro 1397:
1398: if ( !g ) {
1399: *rp = 0;
1400: return 1;
1401: }
1402: sugar = SG(g);
1403: n = NV(g);
1.34 noro 1404: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1405: bucket = create_pbucket();
1.31 noro 1406: add_pbucket(mod,bucket,g);
1.25 noro 1407: d = 0;
1.61 noro 1408: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1409: while ( 1 ) {
1.26 noro 1410: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1411: if ( hindex < 0 ) {
1.77 ! noro 1412: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1413: if ( d ) SG(d) = sugar;
1.25 noro 1414: *rp = d;
1415: return 1;
1416: }
1417: g = bucket->body[hindex];
1.63 noro 1418: index = ndl_find_reducer(HDL(g));
1.25 noro 1419: if ( index >= 0 ) {
1.77 ! noro 1420: count++;
! 1421: if ( !d ) hcount++;
1.25 noro 1422: h = nd_psh[index];
1423: ndl_sub(HDL(g),DL(h),DL(mul));
1424: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1425: nd_free(d);
1.25 noro 1426: free_pbucket(bucket);
1427: *rp = 0;
1428: return 0;
1429: }
1.53 noro 1430: p = ps[index];
1.71 noro 1431: if ( mod == -1 )
1432: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1433: else if ( mod ) {
1.25 noro 1434: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1435: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1436: } else {
1437: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1438: chsgnq(cg,&CQ(mul));
1.26 noro 1439: nd_mul_c_q(d,cred);
1440: mulq_pbucket(bucket,cred);
1441: g = bucket->body[hindex];
1.28 noro 1442: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1443: }
1.55 noro 1444: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1445: bucket->body[hindex] = nd_remove_head(g);
1446: red = nd_remove_head(red);
1.31 noro 1447: add_pbucket(mod,bucket,red);
1.34 noro 1448: psugar = SG(p)+TD(DL(mul));
1449: sugar = MAX(sugar,psugar);
1.28 noro 1450: if ( !mod && hmag && (gmag > hmag) ) {
1451: g = normalize_pbucket(mod,bucket);
1452: if ( !g ) {
1.34 noro 1453: if ( d ) SG(d) = sugar;
1.28 noro 1454: *rp = d;
1455: return 1;
1456: }
1457: nd_removecont2(d,g);
1458: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1459: add_pbucket(mod,bucket,g);
1.28 noro 1460: }
1.25 noro 1461: } else if ( !full ) {
1462: g = normalize_pbucket(mod,bucket);
1.34 noro 1463: if ( g ) SG(g) = sugar;
1.25 noro 1464: *rp = g;
1465: return 1;
1466: } else {
1467: m = BDY(g);
1468: if ( NEXT(m) ) {
1.34 noro 1469: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1470: } else {
1471: FREEND(g); g = 0;
1472: }
1473: bucket->body[hindex] = g;
1474: NEXT(m) = 0;
1475: if ( d ) {
1.34 noro 1476: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1477: } else {
1.34 noro 1478: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1479: }
1480: }
1481: }
1482: }
1.27 noro 1483:
1.61 noro 1484: /* input : list of NDV, cand : list of NDV */
1.28 noro 1485:
1.61 noro 1486: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1487: {
1488: int n,i,stat;
1489: ND nf,d;
1.61 noro 1490: NDV r;
1.45 noro 1491: NODE t,s;
1.69 noro 1492: union oNDC dn;
1.45 noro 1493:
1.61 noro 1494: ndv_setup(0,0,cand);
1.31 noro 1495: n = length(cand);
1.28 noro 1496:
1497: /* membercheck : list is a subset of Id(cand) ? */
1498: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1499: again:
1.61 noro 1500: if ( nd_bpe > obpe )
1501: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1502: else
1503: r = (NDV)BDY(t);
1504: d = ndvtond(0,r);
1.69 noro 1505: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1506: if ( !stat ) {
1507: nd_reconstruct(0,0,0);
1508: goto again;
1509: } else if ( nf ) return 0;
1.71 noro 1510: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1511: }
1.71 noro 1512: if ( DP_Print ) { printf("\n"); }
1.28 noro 1513: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 ! noro 1514: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1515: /* XXX */
1.23 noro 1516: return 1;
1517: }
1.1 noro 1518:
1519: ND nd_remove_head(ND p)
1520: {
1521: NM m;
1522:
1523: m = BDY(p);
1524: if ( !NEXT(m) ) {
1.34 noro 1525: FREEND(p); p = 0;
1.31 noro 1526: } else {
1.34 noro 1527: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1528: }
1.1 noro 1529: FREENM(m);
1530: return p;
1531: }
1532:
1.69 noro 1533: ND nd_separate_head(ND p,ND *head)
1534: {
1535: NM m,m0;
1536: ND r;
1537:
1538: m = BDY(p);
1539: if ( !NEXT(m) ) {
1540: *head = p; p = 0;
1541: } else {
1542: m0 = m;
1543: BDY(p) = NEXT(m); LEN(p)--;
1544: NEXT(m0) = 0;
1545: MKND(NV(p),m0,1,r);
1546: *head = r;
1547: }
1548: return p;
1549: }
1550:
1.1 noro 1551: PGeoBucket create_pbucket()
1552: {
1553: PGeoBucket g;
1554:
1555: g = CALLOC(1,sizeof(struct oPGeoBucket));
1556: g->m = -1;
1557: return g;
1558: }
1559:
1.25 noro 1560: void free_pbucket(PGeoBucket b) {
1561: int i;
1562:
1.26 noro 1563: for ( i = 0; i <= b->m; i++ )
1.25 noro 1564: if ( b->body[i] ) {
1565: nd_free(b->body[i]);
1566: b->body[i] = 0;
1567: }
1568: GC_free(b);
1569: }
1570:
1.63 noro 1571: void add_pbucket_symbolic(PGeoBucket g,ND d)
1572: {
1573: int l,i,k,m;
1574:
1575: if ( !d )
1576: return;
1577: l = LEN(d);
1578: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1579: /* 2^(k-1) < l <= 2^k (=m) */
1580: d = nd_merge(g->body[k],d);
1581: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1582: g->body[k] = 0;
1583: d = nd_merge(g->body[k+1],d);
1584: }
1585: g->body[k] = d;
1586: g->m = MAX(g->m,k);
1587: }
1588:
1.31 noro 1589: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1590: {
1.31 noro 1591: int l,i,k,m;
1.1 noro 1592:
1.31 noro 1593: if ( !d )
1594: return;
1595: l = LEN(d);
1.29 noro 1596: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1597: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1598: d = nd_add(mod,g->body[k],d);
1599: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1600: g->body[k] = 0;
1.31 noro 1601: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1602: }
1603: g->body[k] = d;
1604: g->m = MAX(g->m,k);
1605: }
1606:
1.26 noro 1607: void mulq_pbucket(PGeoBucket g,Q c)
1608: {
1609: int k;
1610:
1611: for ( k = 0; k <= g->m; k++ )
1612: nd_mul_c_q(g->body[k],c);
1613: }
1614:
1.63 noro 1615: NM remove_head_pbucket_symbolic(PGeoBucket g)
1616: {
1617: int j,i,k,c;
1618: NM head;
1619:
1620: k = g->m;
1621: j = -1;
1622: for ( i = 0; i <= k; i++ ) {
1623: if ( !g->body[i] ) continue;
1624: if ( j < 0 ) j = i;
1625: else {
1626: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1627: if ( c > 0 )
1628: j = i;
1629: else if ( c == 0 )
1630: g->body[i] = nd_remove_head(g->body[i]);
1631: }
1632: }
1633: if ( j < 0 ) return 0;
1634: else {
1635: head = BDY(g->body[j]);
1636: if ( !NEXT(head) ) {
1637: FREEND(g->body[j]);
1638: g->body[j] = 0;
1639: } else {
1640: BDY(g->body[j]) = NEXT(head);
1641: LEN(g->body[j])--;
1642: }
1643: return head;
1644: }
1645: }
1646:
1.19 noro 1647: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1648: {
1649: int j,i,c,k,nv,sum;
1.61 noro 1650: UINT *di,*dj;
1.1 noro 1651: ND gi,gj;
1652:
1653: k = g->m;
1654: while ( 1 ) {
1655: j = -1;
1656: for ( i = 0; i <= k; i++ ) {
1657: if ( !(gi = g->body[i]) )
1658: continue;
1659: if ( j < 0 ) {
1660: j = i;
1661: gj = g->body[j];
1662: dj = HDL(gj);
1.14 noro 1663: sum = HCM(gj);
1.1 noro 1664: } else {
1.34 noro 1665: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1666: if ( c > 0 ) {
1.34 noro 1667: if ( sum ) HCM(gj) = sum;
1668: else g->body[j] = nd_remove_head(gj);
1.1 noro 1669: j = i;
1670: gj = g->body[j];
1671: dj = HDL(gj);
1.14 noro 1672: sum = HCM(gj);
1.1 noro 1673: } else if ( c == 0 ) {
1.71 noro 1674: if ( mod == -1 )
1675: sum = _addsf(sum,HCM(gi));
1676: else {
1677: sum = sum+HCM(gi)-mod;
1678: if ( sum < 0 ) sum += mod;
1679: }
1.1 noro 1680: g->body[i] = nd_remove_head(gi);
1681: }
1682: }
1683: }
1.34 noro 1684: if ( j < 0 ) return -1;
1.1 noro 1685: else if ( sum ) {
1.14 noro 1686: HCM(gj) = sum;
1.26 noro 1687: return j;
1.31 noro 1688: } else
1.26 noro 1689: g->body[j] = nd_remove_head(gj);
1690: }
1691: }
1692:
1693: int head_pbucket_q(PGeoBucket g)
1694: {
1695: int j,i,c,k,nv;
1696: Q sum,t;
1697: ND gi,gj;
1698:
1699: k = g->m;
1700: while ( 1 ) {
1701: j = -1;
1702: for ( i = 0; i <= k; i++ ) {
1.34 noro 1703: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1704: if ( j < 0 ) {
1705: j = i;
1706: gj = g->body[j];
1707: sum = HCQ(gj);
1708: } else {
1709: nv = NV(gi);
1.34 noro 1710: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1711: if ( c > 0 ) {
1.34 noro 1712: if ( sum ) HCQ(gj) = sum;
1713: else g->body[j] = nd_remove_head(gj);
1.26 noro 1714: j = i;
1715: gj = g->body[j];
1716: sum = HCQ(gj);
1717: } else if ( c == 0 ) {
1718: addq(sum,HCQ(gi),&t);
1719: sum = t;
1720: g->body[i] = nd_remove_head(gi);
1721: }
1722: }
1723: }
1.34 noro 1724: if ( j < 0 ) return -1;
1.26 noro 1725: else if ( sum ) {
1726: HCQ(gj) = sum;
1.1 noro 1727: return j;
1.31 noro 1728: } else
1.1 noro 1729: g->body[j] = nd_remove_head(gj);
1730: }
1731: }
1732:
1.25 noro 1733: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1734: {
1.31 noro 1735: int i;
1.1 noro 1736: ND r,t;
1737:
1738: r = 0;
1.28 noro 1739: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1740: r = nd_add(mod,r,g->body[i]);
1.28 noro 1741: g->body[i] = 0;
1742: }
1743: g->m = -1;
1.1 noro 1744: return r;
1745: }
1746:
1.27 noro 1747: /* return value = 0 => input is not a GB */
1748:
1.77 ! noro 1749: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1750: {
1751: int i,nh,sugar,stat;
1.23 noro 1752: NODE r,g,t;
1.1 noro 1753: ND_pairs d;
1754: ND_pairs l;
1.77 ! noro 1755: ND h,nf,s,head;
1.63 noro 1756: NDV nfv;
1.77 ! noro 1757: Q q;
1.69 noro 1758: union oNDC dn;
1.1 noro 1759:
1.23 noro 1760: g = 0; d = 0;
1761: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1762: d = update_pairs(d,g,i);
1763: g = update_base(g,i);
1764: }
1765: sugar = 0;
1766: while ( d ) {
1767: again:
1768: l = nd_minp(d,&d);
1.14 noro 1769: if ( SG(l) != sugar ) {
1.77 ! noro 1770: if ( ishomo ) {
! 1771: for ( i = nd_psn-1; SG(nd_ps[i]) == sugar; i-- ) {
! 1772: s = ndvtond(m,nd_ps[i]);
! 1773: s = nd_separate_head(s,&head);
! 1774: nd_nf(m,s,nd_ps,1,&dn,&nf);
! 1775: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
! 1776: nf = nd_add(m,head,nf);
! 1777: ndv_free(nd_ps[i]);
! 1778: nd_removecont(m,nf);
! 1779: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
! 1780: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
! 1781: }
! 1782: }
1.14 noro 1783: sugar = SG(l);
1.71 noro 1784: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1785: }
1.53 noro 1786: stat = nd_sp(m,0,l,&h);
1.1 noro 1787: if ( !stat ) {
1788: NEXT(l) = d; d = l;
1.20 noro 1789: d = nd_reconstruct(m,0,d);
1.1 noro 1790: goto again;
1791: }
1.41 noro 1792: #if USE_GEOBUCKET
1.69 noro 1793: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1794: #else
1.69 noro 1795: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1796: #endif
1.1 noro 1797: if ( !stat ) {
1798: NEXT(l) = d; d = l;
1.20 noro 1799: d = nd_reconstruct(m,0,d);
1.1 noro 1800: goto again;
1801: } else if ( nf ) {
1.27 noro 1802: if ( checkonly ) return 0;
1.71 noro 1803: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1804: nd_removecont(m,nf);
1805: nfv = ndtondv(m,nf); nd_free(nf);
1.77 ! noro 1806: nh = ndv_newps(m,nfv,0);
1.1 noro 1807: d = update_pairs(d,g,nh);
1808: g = update_base(g,nh);
1809: FREENDP(l);
1810: } else {
1.71 noro 1811: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1812: FREENDP(l);
1813: }
1814: }
1.77 ! noro 1815: if ( nd_demand )
! 1816: for ( t = g; t; t = NEXT(t) )
! 1817: BDY(t) = (pointer)ndv_load((int)BDY(t));
! 1818: else
! 1819: for ( t = g; t; t = NEXT(t) )
! 1820: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1821: return g;
1822: }
1823:
1.23 noro 1824: NODE nd_gb_trace(int m)
1.20 noro 1825: {
1826: int i,nh,sugar,stat;
1.23 noro 1827: NODE r,g,t;
1.20 noro 1828: ND_pairs d;
1829: ND_pairs l;
1830: ND h,nf,nfq;
1.63 noro 1831: NDV nfv,nfqv;
1.69 noro 1832: union oNDC dn;
1.20 noro 1833:
1.23 noro 1834: g = 0; d = 0;
1835: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1836: d = update_pairs(d,g,i);
1837: g = update_base(g,i);
1838: }
1839: sugar = 0;
1840: while ( d ) {
1841: again:
1842: l = nd_minp(d,&d);
1843: if ( SG(l) != sugar ) {
1844: sugar = SG(l);
1.71 noro 1845: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1846: }
1.53 noro 1847: stat = nd_sp(m,0,l,&h);
1.20 noro 1848: if ( !stat ) {
1849: NEXT(l) = d; d = l;
1850: d = nd_reconstruct(m,1,d);
1851: goto again;
1852: }
1.41 noro 1853: #if USE_GEOBUCKET
1.53 noro 1854: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1855: #else
1.69 noro 1856: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1857: #endif
1.20 noro 1858: if ( !stat ) {
1859: NEXT(l) = d; d = l;
1860: d = nd_reconstruct(m,1,d);
1861: goto again;
1862: } else if ( nf ) {
1.77 ! noro 1863: if ( nd_demand ) {
! 1864: nfqv = ndv_load(nd_psn);
! 1865: nfq = ndvtond(0,nfqv);
! 1866: } else
! 1867: nfq = 0;
! 1868: if ( !nfq ) {
! 1869: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
! 1870: NEXT(l) = d; d = l;
! 1871: d = nd_reconstruct(m,1,d);
! 1872: goto again;
! 1873: }
1.70 noro 1874: }
1.20 noro 1875: if ( nfq ) {
1.70 noro 1876: /* m|HC(nfq) => failure */
1.63 noro 1877: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1878:
1.71 noro 1879: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1880: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1881: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 ! noro 1882: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1883: d = update_pairs(d,g,nh);
1884: g = update_base(g,nh);
1885: } else {
1.71 noro 1886: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1887: }
1888: } else {
1.71 noro 1889: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1890: }
1891: FREENDP(l);
1892: }
1.77 ! noro 1893: if ( nd_demand )
! 1894: for ( t = g; t; t = NEXT(t) )
! 1895: BDY(t) = (pointer)ndv_load((int)BDY(t));
! 1896: else
! 1897: for ( t = g; t; t = NEXT(t) )
! 1898: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1899: return g;
1900: }
1901:
1.23 noro 1902: int ndv_compare(NDV *p1,NDV *p2)
1903: {
1.34 noro 1904: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1905: }
1906:
1907: int ndv_compare_rev(NDV *p1,NDV *p2)
1908: {
1.34 noro 1909: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1910: }
1911:
1.61 noro 1912: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1913: {
1.69 noro 1914: int i,n,stat;
1915: ND nf,g,head;
1.23 noro 1916: NODE t,a0,a;
1.69 noro 1917: union oNDC dn;
1918: NDV *w;
1919: Q q;
1.23 noro 1920:
1.69 noro 1921: n = length(f);
1922: #if 0
1923: w = (NDV *)ALLOCA(n*sizeof(NDV));
1924: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1925: qsort(w,n,sizeof(NDV),
1926: (int (*)(const void *,const void *))ndv_compare);
1927: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1928: #endif
1929: ndv_setup(m,0,f);
1930: for ( i = 0; i < n; ) {
1931: g = ndvtond(m,nd_ps[i]);
1932: g = nd_separate_head(g,&head);
1933: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1934: if ( !stat )
1.69 noro 1935: nd_reconstruct(m,0,0);
1936: else {
1.71 noro 1937: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1938: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1939: nf = nd_add(m,head,nf);
1940: ndv_free(nd_ps[i]);
1.24 noro 1941: nd_removecont(m,nf);
1.69 noro 1942: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1943: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1944: i++;
1.23 noro 1945: }
1946: }
1.71 noro 1947: if ( DP_Print ) { printf("\n"); }
1.23 noro 1948: for ( a0 = 0, i = 0; i < n; i++ ) {
1949: NEXTNODE(a0,a);
1.69 noro 1950: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1951: }
1952: NEXT(a) = 0;
1953: return a0;
1954: }
1955:
1.1 noro 1956: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1957: {
1958: ND_pairs d1,nd,cur,head,prev,remove;
1959:
1960: if ( !g ) return d;
1961: d = crit_B(d,t);
1962: d1 = nd_newpairs(g,t);
1963: d1 = crit_M(d1);
1964: d1 = crit_F(d1);
1.55 noro 1965: if ( do_weyl )
1966: head = d1;
1967: else {
1968: prev = 0; cur = head = d1;
1969: while ( cur ) {
1970: if ( crit_2( cur->i1,cur->i2 ) ) {
1971: remove = cur;
1972: if ( !prev ) head = cur = NEXT(cur);
1973: else cur = NEXT(prev) = NEXT(cur);
1974: FREENDP(remove);
1975: } else {
1976: prev = cur; cur = NEXT(cur);
1977: }
1.1 noro 1978: }
1979: }
1980: if ( !d )
1981: return head;
1982: else {
1983: nd = d;
1.34 noro 1984: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1985: NEXT(nd) = head;
1986: return d;
1987: }
1988: }
1989:
1990: ND_pairs nd_newpairs( NODE g, int t )
1991: {
1992: NODE h;
1.61 noro 1993: UINT *dl;
1.34 noro 1994: int ts,s;
1.1 noro 1995: ND_pairs r,r0;
1996:
1.20 noro 1997: dl = DL(nd_psh[t]);
1.34 noro 1998: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1999: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
2000: NEXTND_pairs(r0,r);
2001: r->i1 = (int)BDY(h);
2002: r->i2 = t;
1.20 noro 2003: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 2004: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
2005: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 2006: }
2007: NEXT(r) = 0;
2008: return r0;
2009: }
2010:
2011: ND_pairs crit_B( ND_pairs d, int s )
2012: {
2013: ND_pairs cur,head,prev,remove;
1.61 noro 2014: UINT *t,*tl,*lcm;
1.1 noro 2015: int td,tdl;
2016:
2017: if ( !d ) return 0;
1.20 noro 2018: t = DL(nd_psh[s]);
1.1 noro 2019: prev = 0;
2020: head = cur = d;
1.61 noro 2021: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 2022: while ( cur ) {
2023: tl = cur->lcm;
2024: if ( ndl_reducible(tl,t)
1.20 noro 2025: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
2026: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 2027: remove = cur;
2028: if ( !prev ) {
2029: head = cur = NEXT(cur);
2030: } else {
2031: cur = NEXT(prev) = NEXT(cur);
2032: }
2033: FREENDP(remove);
2034: } else {
1.34 noro 2035: prev = cur; cur = NEXT(cur);
1.1 noro 2036: }
2037: }
2038: return head;
2039: }
2040:
2041: ND_pairs crit_M( ND_pairs d1 )
2042: {
2043: ND_pairs e,d2,d3,dd,p;
1.61 noro 2044: UINT *id,*jd;
1.1 noro 2045:
2046: for ( dd = 0, e = d1; e; e = d3 ) {
2047: if ( !(d2 = NEXT(e)) ) {
2048: NEXT(e) = dd;
2049: return e;
2050: }
1.34 noro 2051: id = LCM(e);
1.1 noro 2052: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2053: p = NEXT(d2);
2054: jd = LCM(d2);
2055: if ( ndl_equal(jd,id) )
2056: ;
2057: else if ( TD(jd) > TD(id) )
1.1 noro 2058: if ( ndl_reducible(jd,id) ) continue;
2059: else ;
1.34 noro 2060: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2061: NEXT(d2) = d3;
2062: d3 = d2;
2063: }
2064: NEXT(e) = dd;
2065: dd = e;
2066: continue;
2067: /**/
2068: delit: NEXT(d2) = d3;
2069: d3 = d2;
2070: for ( ; p; p = d2 ) {
2071: d2 = NEXT(p);
2072: NEXT(p) = d3;
2073: d3 = p;
2074: }
2075: FREENDP(e);
2076: }
2077: return dd;
2078: }
2079:
2080: ND_pairs crit_F( ND_pairs d1 )
2081: {
2082: ND_pairs rest, head,remove;
2083: ND_pairs last, p, r, w;
2084: int s;
2085:
2086: for ( head = last = 0, p = d1; NEXT(p); ) {
2087: r = w = equivalent_pairs(p,&rest);
1.14 noro 2088: s = SG(r);
1.1 noro 2089: w = NEXT(w);
2090: while ( w ) {
2091: if ( crit_2(w->i1,w->i2) ) {
2092: r = w;
2093: w = NEXT(w);
2094: while ( w ) {
2095: remove = w;
2096: w = NEXT(w);
2097: FREENDP(remove);
2098: }
2099: break;
1.14 noro 2100: } else if ( SG(w) < s ) {
1.1 noro 2101: FREENDP(r);
2102: r = w;
1.14 noro 2103: s = SG(r);
1.1 noro 2104: w = NEXT(w);
2105: } else {
2106: remove = w;
2107: w = NEXT(w);
2108: FREENDP(remove);
2109: }
2110: }
2111: if ( last ) NEXT(last) = r;
2112: else head = r;
2113: NEXT(last = r) = 0;
2114: p = rest;
2115: if ( !p ) return head;
2116: }
2117: if ( !last ) return p;
2118: NEXT(last) = p;
2119: return head;
2120: }
2121:
2122: int crit_2( int dp1, int dp2 )
2123: {
1.20 noro 2124: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2125: }
2126:
1.40 noro 2127: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2128: {
2129: ND_pairs w,p,r,s;
1.61 noro 2130: UINT *d;
1.1 noro 2131:
2132: w = d1;
1.34 noro 2133: d = LCM(w);
1.1 noro 2134: s = NEXT(w);
2135: NEXT(w) = 0;
2136: for ( r = 0; s; s = p ) {
2137: p = NEXT(s);
1.34 noro 2138: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2139: NEXT(s) = w; w = s;
1.1 noro 2140: } else {
1.39 noro 2141: NEXT(s) = r; r = s;
1.1 noro 2142: }
2143: }
2144: *prest = r;
2145: return w;
2146: }
2147:
2148: NODE update_base(NODE nd,int ndp)
2149: {
1.61 noro 2150: UINT *dl, *dln;
1.1 noro 2151: NODE last, p, head;
2152:
1.20 noro 2153: dl = DL(nd_psh[ndp]);
1.1 noro 2154: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2155: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2156: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2157: p = NEXT(p);
2158: if ( last ) NEXT(last) = p;
2159: } else {
2160: if ( !last ) head = p;
2161: p = NEXT(last = p);
2162: }
2163: }
2164: head = append_one(head,ndp);
2165: return head;
2166: }
2167:
2168: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2169: {
2170: ND_pairs m,ml,p,l;
1.61 noro 2171: UINT *lcm;
1.33 noro 2172: int s,td,len,tlen,c,c1;
1.1 noro 2173:
2174: if ( !(p = NEXT(m = d)) ) {
2175: *prest = p;
2176: NEXT(m) = 0;
2177: return m;
2178: }
1.14 noro 2179: s = SG(m);
1.77 ! noro 2180: if ( !NoSugar ) {
! 2181: for ( ml = 0, l = m; p; p = NEXT(l = p) )
! 2182: if ( (SG(p) < s)
! 2183: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
! 2184: ml = l; m = p; s = SG(m);
! 2185: }
! 2186: } else {
! 2187: for ( ml = 0, l = m; p; p = NEXT(l = p) )
! 2188: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
! 2189: ml = l; m = p; s = SG(m);
! 2190: }
! 2191: }
1.1 noro 2192: if ( !ml ) *prest = NEXT(m);
2193: else {
2194: NEXT(ml) = NEXT(m);
2195: *prest = d;
2196: }
2197: NEXT(m) = 0;
2198: return m;
2199: }
2200:
1.63 noro 2201: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2202: {
1.69 noro 2203: int msugar,i;
1.63 noro 2204: ND_pairs t,dm0,dm,dr0,dr;
2205:
2206: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2207: if ( SG(t) < msugar ) msugar = SG(t);
2208: dm0 = 0; dr0 = 0;
1.69 noro 2209: for ( i = 0, t = d; t; t = NEXT(t) )
2210: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2211: if ( dm0 ) NEXT(dm) = t;
2212: else dm0 = t;
2213: dm = t;
1.69 noro 2214: i++;
1.63 noro 2215: } else {
2216: if ( dr0 ) NEXT(dr) = t;
2217: else dr0 = t;
2218: dr = t;
2219: }
2220: NEXT(dm) = 0;
2221: if ( dr0 ) NEXT(dr) = 0;
2222: *prest = dr0;
2223: return dm0;
2224: }
2225:
1.77 ! noro 2226: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2227: {
1.3 noro 2228: int len;
1.13 noro 2229: RHist r;
1.20 noro 2230: NDV b;
1.3 noro 2231:
1.1 noro 2232: if ( nd_psn == nd_pslen ) {
2233: nd_pslen *= 2;
1.11 noro 2234: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2235: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2236: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2237: nd_bound = (UINT **)
2238: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2239: }
1.39 noro 2240: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2241: nd_ps[nd_psn] = a;
1.39 noro 2242: if ( aq ) {
1.63 noro 2243: nd_ps_trace[nd_psn] = aq;
2244: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2245: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2246: } else {
1.63 noro 2247: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2248: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2249: }
1.77 ! noro 2250: if ( nd_demand ) {
! 2251: if ( aq ) {
! 2252: ndv_save(nd_ps_trace[nd_psn],nd_psn);
! 2253: nd_ps_trace[nd_psn] = 0;
! 2254: } else {
! 2255: ndv_save(nd_ps[nd_psn],nd_psn);
! 2256: nd_ps[nd_psn] = 0;
! 2257: }
! 2258: }
1.1 noro 2259: return nd_psn++;
2260: }
2261:
1.61 noro 2262: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2263: {
1.5 noro 2264: int i,j,td,len,max;
1.1 noro 2265: NODE s,s0,f0;
1.61 noro 2266: UINT *d;
1.13 noro 2267: RHist r;
1.69 noro 2268: NDV *w;
1.61 noro 2269: NDV a,am;
1.11 noro 2270:
2271: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2272:
1.69 noro 2273: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2274: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2275: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2276: qsort(w,nd_psn,sizeof(NDV),
2277: (int (*)(const void *,const void *))ndv_compare);
2278: nd_pslen = 2*nd_psn;
1.11 noro 2279: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2280: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2281: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2282: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2283:
1.11 noro 2284: if ( !nd_red )
1.13 noro 2285: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2286: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2287: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2288: if ( trace ) {
1.69 noro 2289: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2290: ndv_removecont(0,a);
2291: am = nd_ps[i] = ndv_dup(mod,a);
2292: ndv_mod(mod,am);
2293: ndv_removecont(mod,am);
2294: } else {
1.69 noro 2295: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2296: ndv_removecont(mod,a);
1.39 noro 2297: }
1.61 noro 2298: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2299: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2300: nd_psh[i] = r;
1.77 ! noro 2301: if ( nd_demand ) {
! 2302: if ( trace ) {
! 2303: ndv_save(nd_ps_trace[i],i);
! 2304: nd_ps_trace[i] = 0;
! 2305: } else {
! 2306: ndv_save(nd_ps[i],i);
! 2307: nd_ps[i] = 0;
! 2308: }
! 2309: }
1.20 noro 2310: }
2311: }
2312:
1.63 noro 2313: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2314: {
1.61 noro 2315: VL tv,fv,vv,vc;
1.1 noro 2316: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2317: int e,max,nvar;
1.69 noro 2318: NDV b;
1.77 ! noro 2319: int ishomo;
1.1 noro 2320:
1.74 noro 2321: ndv_alloc = 0;
1.1 noro 2322: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2323: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.32 noro 2324: nd_init_ord(ord);
1.61 noro 2325: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2326: for ( tv = vv; tv; tv = NEXT(tv) ) {
2327: e = getdeg(tv->v,(P)BDY(t));
2328: max = MAX(e,max);
2329: }
2330: nd_setup_parameters(nvar,max);
1.77 ! noro 2331: ishomo = 1;
1.1 noro 2332: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2333: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 ! noro 2334: if ( ishomo )
! 2335: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2336: if ( m ) ndv_mod(m,b);
1.61 noro 2337: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2338: }
2339: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2340: ndv_setup(m,0,fd0);
1.77 ! noro 2341: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.61 noro 2342: x = ndv_reducebase(x);
2343: x = ndv_reduceall(m,x);
1.23 noro 2344: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2345: NEXTNODE(r0,r);
1.61 noro 2346: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2347: }
2348: if ( r0 ) NEXT(r) = 0;
2349: MKLIST(*rp,r0);
1.74 noro 2350: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2351: }
2352:
1.52 noro 2353: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2354: {
2355: struct order_spec ord1;
1.61 noro 2356: VL tv,fv,vv,vc;
1.27 noro 2357: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2358: int m,nocheck,nvar,mindex,e,max;
2359: NDV c;
2360: NMV a;
1.27 noro 2361: P p;
1.61 noro 2362: EPOS oepos;
2363: int obpe,oadv,wmax,i,len,cbpe;
1.20 noro 2364:
2365: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2366: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.52 noro 2367: nocheck = 0;
2368: mindex = 0;
2369:
1.77 ! noro 2370: if ( Demand )
! 2371: nd_demand = 1;
1.52 noro 2372: /* setup modulus */
2373: if ( trace < 0 ) {
2374: trace = -trace;
2375: nocheck = 1;
2376: }
2377: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2378: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2379: for ( tv = vv; tv; tv = NEXT(tv) ) {
2380: e = getdeg(tv->v,(P)BDY(t));
2381: max = MAX(e,max);
1.23 noro 2382: }
1.61 noro 2383: nd_init_ord(ord);
2384: nd_setup_parameters(nvar,max);
2385: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2386: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2387: c = ptondv(CO,vv,(P)BDY(t));
2388: if ( c ) {
2389: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2390: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2391: }
1.61 noro 2392: }
2393: if ( in0 ) NEXT(in) = 0;
2394: if ( fd0 ) NEXT(fd) = 0;
2395: if ( homo ) {
2396: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2397: c = (NDV)BDY(t); len = LEN(c);
2398: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2399: wmax = MAX(TD(DL(a)),wmax);
2400: }
2401: homogenize_order(ord,nvar,&ord1);
2402: nd_init_ord(&ord1);
2403: nd_setup_parameters(nvar+1,wmax);
2404: for ( t = fd0; t; t = NEXT(t) )
2405: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2406: }
1.52 noro 2407: while ( 1 ) {
1.77 ! noro 2408: if ( Demand )
! 2409: nd_demand = 1;
1.61 noro 2410: ndv_setup(m,1,fd0);
1.27 noro 2411: cand = nd_gb_trace(m);
1.52 noro 2412: if ( !cand ) {
2413: /* failure */
1.61 noro 2414: if ( trace > 1 ) { *rp = 0; return; }
2415: else m = get_lprime(++mindex);
1.52 noro 2416: continue;
2417: }
1.27 noro 2418: if ( homo ) {
2419: /* dehomogenization */
1.61 noro 2420: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2421: nd_init_ord(ord);
1.61 noro 2422: nd_setup_parameters(nvar,0);
1.27 noro 2423: }
1.77 ! noro 2424: nd_demand = 0;
1.61 noro 2425: cand = ndv_reducebase(cand);
2426: cand = ndv_reduceall(0,cand);
2427: if ( nocheck )
2428: break;
2429: cbpe = nd_bpe;
2430: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2431: /* success */
2432: break;
2433: else if ( trace > 1 ) {
2434: /* failure */
2435: *rp = 0; return;
1.61 noro 2436: } else {
1.52 noro 2437: /* try the next modulus */
2438: m = get_lprime(++mindex);
1.61 noro 2439: /* reset the parameters */
2440: if ( homo ) {
2441: nd_init_ord(&ord1);
2442: nd_setup_parameters(nvar+1,wmax);
2443: } else {
2444: nd_init_ord(ord);
2445: nd_setup_parameters(nvar,max);
2446: }
2447: }
1.52 noro 2448: }
1.27 noro 2449: /* dp->p */
1.61 noro 2450: nd_bpe = cbpe;
1.73 noro 2451: nd_setup_parameters(nd_nvar,0);
1.61 noro 2452: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2453: MKLIST(*rp,cand);
1.1 noro 2454: }
2455:
1.61 noro 2456: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2457: {
1.61 noro 2458: UINT *d;
1.57 noro 2459: int i,j,l,s,ord_l;
1.43 noro 2460: struct order_pair *op;
1.1 noro 2461:
2462: d = dl->d;
1.41 noro 2463: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2464: if ( nd_blockmask ) {
2465: l = nd_blockmask->n;
2466: op = nd_blockmask->order_pair;
2467: for ( j = 0, s = 0; j < l; j++ ) {
2468: ord_l = op[j].length;
1.57 noro 2469: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2470: }
2471: TD(r) = ndl_weight(r);
1.61 noro 2472: ndl_weight_mask(r);
1.43 noro 2473: } else {
1.56 noro 2474: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2475: TD(r) = ndl_weight(r);
2476: }
1.1 noro 2477: }
2478:
1.61 noro 2479: DL ndltodl(int n,UINT *ndl)
1.1 noro 2480: {
2481: DL dl;
2482: int *d;
1.57 noro 2483: int i,j,l,s,ord_l;
1.43 noro 2484: struct order_pair *op;
1.1 noro 2485:
2486: NEWDL(dl,n);
1.34 noro 2487: dl->td = TD(ndl);
1.1 noro 2488: d = dl->d;
1.43 noro 2489: if ( nd_blockmask ) {
2490: l = nd_blockmask->n;
2491: op = nd_blockmask->order_pair;
2492: for ( j = 0, s = 0; j < l; j++ ) {
2493: ord_l = op[j].length;
1.57 noro 2494: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2495: }
2496: } else {
1.56 noro 2497: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2498: }
1.1 noro 2499: return dl;
2500: }
2501:
1.61 noro 2502: void ndl_print(UINT *dl)
1.1 noro 2503: {
2504: int n;
1.57 noro 2505: int i,j,l,ord_l,s,s0;
1.43 noro 2506: struct order_pair *op;
1.1 noro 2507:
2508: n = nd_nvar;
2509: printf("<<");
1.43 noro 2510: if ( nd_blockmask ) {
2511: l = nd_blockmask->n;
2512: op = nd_blockmask->order_pair;
2513: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2514: ord_l = op[j].length;
1.57 noro 2515: for ( i = 0; i < ord_l; i++, s++ )
2516: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2517: }
2518: } else {
1.56 noro 2519: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2520: }
1.1 noro 2521: printf(">>");
2522: }
2523:
2524: void nd_print(ND p)
2525: {
2526: NM m;
2527:
2528: if ( !p )
2529: printf("0\n");
2530: else {
2531: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2532: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2533: else printf("+%d*",CM(m));
1.14 noro 2534: ndl_print(DL(m));
1.1 noro 2535: }
2536: printf("\n");
2537: }
2538: }
2539:
1.16 noro 2540: void nd_print_q(ND p)
2541: {
2542: NM m;
2543:
2544: if ( !p )
2545: printf("0\n");
2546: else {
2547: for ( m = BDY(p); m; m = NEXT(m) ) {
2548: printf("+");
1.74 noro 2549: printexpr(CO,(Obj)CQ(m));
1.16 noro 2550: printf("*");
2551: ndl_print(DL(m));
2552: }
2553: printf("\n");
2554: }
2555: }
2556:
1.1 noro 2557: void ndp_print(ND_pairs d)
2558: {
2559: ND_pairs t;
2560:
1.34 noro 2561: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2562: printf("\n");
2563: }
2564:
1.20 noro 2565: void nd_removecont(int mod,ND p)
1.16 noro 2566: {
2567: int i,n;
2568: Q *w;
2569: Q dvr,t;
2570: NM m;
1.21 noro 2571: struct oVECT v;
2572: N q,r;
1.16 noro 2573:
1.71 noro 2574: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2575: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2576: else {
2577: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2578: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2579: v.len = n;
2580: v.body = (pointer *)w;
1.34 noro 2581: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2582: removecont_array(w,n);
2583: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2584: }
2585: }
2586:
1.21 noro 2587: void nd_removecont2(ND p1,ND p2)
2588: {
2589: int i,n1,n2,n;
2590: Q *w;
2591: Q dvr,t;
2592: NM m;
2593: struct oVECT v;
2594: N q,r;
2595:
2596: if ( !p1 ) {
2597: nd_removecont(0,p2); return;
2598: } else if ( !p2 ) {
2599: nd_removecont(0,p1); return;
2600: }
2601: n1 = nd_length(p1);
2602: n2 = nd_length(p2);
2603: n = n1+n2;
2604: w = (Q *)ALLOCA(n*sizeof(Q));
2605: v.len = n;
2606: v.body = (pointer *)w;
1.34 noro 2607: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2608: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2609: removecont_array(w,n);
2610: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2611: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2612: }
2613:
1.20 noro 2614: void ndv_removecont(int mod,NDV p)
1.16 noro 2615: {
2616: int i,len;
2617: Q *w;
2618: Q dvr,t;
2619: NMV m;
2620:
1.71 noro 2621: if ( mod == -1 )
2622: ndv_mul_c(mod,p,_invsf(HCM(p)));
2623: else if ( mod )
1.20 noro 2624: ndv_mul_c(mod,p,invm(HCM(p),mod));
2625: else {
2626: len = p->len;
2627: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2628: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2629: sortbynm(w,len);
2630: qltozl(w,len,&dvr);
2631: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2632: divq(CQ(m),dvr,&t); CQ(m) = t;
2633: }
1.16 noro 2634: }
1.21 noro 2635: }
2636:
1.61 noro 2637: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2638: {
2639: int len,i,max;
2640: NMV m,mr0,mr,t;
2641:
2642: len = p->len;
2643: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2644: max = MAX(max,TD(DL(m)));
2645: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2646: m = (NMV)((char *)mr0+(len-1)*oadv);
2647: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2648: t = (NMV)ALLOCA(nmv_adv);
2649: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2650: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2651: CQ(mr) = CQ(m);
2652: ndl_copy(DL(t),DL(mr));
2653: }
2654: NV(p)++;
2655: BDY(p) = mr0;
2656: }
2657:
1.45 noro 2658: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2659: {
1.45 noro 2660: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2661: Q *w;
2662: Q dvr,t;
2663: NMV m,r;
2664:
2665: len = p->len;
2666: newnvar = nd_nvar-1;
1.48 noro 2667: newexporigin = nd_get_exporigin(ord);
1.45 noro 2668: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2669: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2670: ndl_dehomogenize(DL(m));
1.23 noro 2671: if ( newwpd != nd_wpd ) {
1.61 noro 2672: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2673: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2674: CQ(r) = CQ(m);
2675: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2676: adj = nd_exporigin-newexporigin;
2677: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2678: }
2679: }
2680: NV(p)--;
2681: }
2682:
1.21 noro 2683: void removecont_array(Q *c,int n)
2684: {
2685: struct oVECT v;
2686: Q d0,d1,a,u,u1,gcd;
2687: int i;
2688: N qn,rn,gn;
2689: Q *q,*r;
2690:
2691: q = (Q *)ALLOCA(n*sizeof(Q));
2692: r = (Q *)ALLOCA(n*sizeof(Q));
2693: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2694: igcdv_estimate(&v,&d0);
2695: for ( i = 0; i < n; i++ ) {
2696: divn(NM(c[i]),NM(d0),&qn,&rn);
2697: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2698: NTOQ(rn,SGN(c[i]),r[i]);
2699: }
1.34 noro 2700: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2701: if ( i < n ) {
2702: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2703: igcdv(&v,&d1);
2704: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2705: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2706: for ( i = 0; i < n; i++ ) {
2707: mulq(a,q[i],&u);
2708: if ( r[i] ) {
2709: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2710: addq(u,u1,&q[i]);
2711: } else
2712: q[i] = u;
2713: }
2714: }
1.34 noro 2715: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2716: }
2717:
1.19 noro 2718: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2719: {
2720: NM m;
2721: int c,c1;
2722:
1.34 noro 2723: if ( !p ) return;
1.71 noro 2724: if ( mod == -1 )
2725: for ( m = BDY(p); m; m = NEXT(m) )
2726: CM(m) = _mulsf(CM(m),mul);
2727: else
2728: for ( m = BDY(p); m; m = NEXT(m) ) {
2729: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2730: }
1.1 noro 2731: }
2732:
1.16 noro 2733: void nd_mul_c_q(ND p,Q mul)
2734: {
2735: NM m;
2736: Q c;
2737:
1.34 noro 2738: if ( !p ) return;
1.16 noro 2739: for ( m = BDY(p); m; m = NEXT(m) ) {
2740: mulq(CQ(m),mul,&c); CQ(m) = c;
2741: }
2742: }
2743:
1.61 noro 2744: void nd_mul_c_p(VL vl,ND p,P mul)
2745: {
2746: NM m;
2747: P c;
2748:
2749: if ( !p ) return;
2750: for ( m = BDY(p); m; m = NEXT(m) ) {
2751: mulp(vl,CP(m),mul,&c); CP(m) = c;
2752: }
2753: }
2754:
1.1 noro 2755: void nd_free(ND p)
2756: {
2757: NM t,s;
2758:
1.34 noro 2759: if ( !p ) return;
1.1 noro 2760: t = BDY(p);
2761: while ( t ) {
2762: s = NEXT(t);
2763: FREENM(t);
2764: t = s;
2765: }
2766: FREEND(p);
2767: }
2768:
1.23 noro 2769: void ndv_free(NDV p)
2770: {
2771: GC_free(BDY(p));
2772: }
2773:
1.61 noro 2774: void nd_append_red(UINT *d,int i)
1.1 noro 2775: {
1.13 noro 2776: RHist m,m0;
1.1 noro 2777: int h;
2778:
1.13 noro 2779: NEWRHist(m);
1.34 noro 2780: h = ndl_hash_value(d);
1.13 noro 2781: m->index = i;
1.14 noro 2782: ndl_copy(d,DL(m));
1.1 noro 2783: NEXT(m) = nd_red[h];
2784: nd_red[h] = m;
2785: }
2786:
1.61 noro 2787: UINT *ndv_compute_bound(NDV p)
1.1 noro 2788: {
1.61 noro 2789: UINT *d1,*d2,*t;
2790: UINT u;
1.57 noro 2791: int i,j,k,l,len,ind;
1.45 noro 2792: NMV m;
1.1 noro 2793:
2794: if ( !p )
2795: return 0;
1.61 noro 2796: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2797: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2798: len = LEN(p);
2799: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2800: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2801: ndl_lcm(DL(m),d1,d2);
1.1 noro 2802: t = d1; d1 = d2; d2 = t;
2803: }
1.12 noro 2804: l = nd_nvar+31;
1.61 noro 2805: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2806: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2807: u = d1[i];
2808: k = (nd_epw-1)*nd_bpe;
2809: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2810: t[ind] = (u>>k)&nd_mask0;
2811: }
2812: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2813: return t;
2814: }
2815:
1.48 noro 2816: int nd_get_exporigin(struct order_spec *ord)
2817: {
1.51 noro 2818: switch ( ord->id ) {
1.41 noro 2819: case 0:
1.48 noro 2820: return 1;
1.41 noro 2821: case 1:
2822: /* block order */
1.43 noro 2823: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2824: return ord->ord.block.length+1;
1.41 noro 2825: case 2:
1.52 noro 2826: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2827: }
1.48 noro 2828: }
2829:
1.61 noro 2830: void nd_setup_parameters(int nvar,int max) {
1.57 noro 2831: int i,j,n,elen,ord_o,ord_l,l,s;
2832: struct order_pair *op;
1.73 noro 2833: int bpe;
1.48 noro 2834:
1.73 noro 2835: if ( !max ) bpe = nd_bpe;
2836: else if ( max < 2 ) bpe = 1;
2837: else if ( max < 4 ) bpe = 2;
2838: else if ( max < 8 ) bpe = 3;
2839: else if ( max < 16 ) bpe = 4;
2840: else if ( max < 32 ) bpe = 5;
2841: else if ( max < 64 ) bpe = 6;
2842: else if ( max < 256 ) bpe = 8;
2843: else if ( max < 1024 ) bpe = 10;
2844: else if ( max < 65536 ) bpe = 16;
2845: else bpe = 32;
2846: if ( bpe != nd_bpe || nvar != nd_nvar )
2847: nd_free_private_storage();
2848: nd_bpe = bpe;
2849: nd_nvar = nvar;
1.61 noro 2850: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2851: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2852:
2853: nd_exporigin = nd_get_exporigin(nd_ord);
1.43 noro 2854: nd_wpd = nd_exporigin+elen;
1.57 noro 2855:
1.1 noro 2856: if ( nd_bpe < 32 ) {
2857: nd_mask0 = (1<<nd_bpe)-1;
2858: } else {
2859: nd_mask0 = 0xffffffff;
2860: }
2861: bzero(nd_mask,sizeof(nd_mask));
2862: nd_mask1 = 0;
2863: for ( i = 0; i < nd_epw; i++ ) {
2864: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2865: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2866: }
1.61 noro 2867: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2868: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2869: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2870: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2871: }
2872:
1.20 noro 2873: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2874: {
1.37 noro 2875: int i,obpe,oadv,h;
1.71 noro 2876: static NM prev_nm_free_list;
2877: static ND_pairs prev_ndp_free_list;
1.13 noro 2878: RHist mr0,mr;
2879: RHist r;
1.37 noro 2880: RHist *old_red;
1.71 noro 2881: ND_pairs s0,s,t;
1.43 noro 2882: EPOS oepos;
1.15 noro 2883:
1.1 noro 2884: obpe = nd_bpe;
1.11 noro 2885: oadv = nmv_adv;
1.43 noro 2886: oepos = nd_epos;
1.63 noro 2887: if ( obpe < 2 ) nd_bpe = 2;
2888: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2889: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2890: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2891: else if ( obpe < 6 ) nd_bpe = 6;
2892: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2893: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2894: else if ( obpe < 16 ) nd_bpe = 16;
2895: else if ( obpe < 32 ) nd_bpe = 32;
2896: else error("nd_reconstruct : exponent too large");
1.5 noro 2897:
1.73 noro 2898: nd_setup_parameters(nd_nvar,0);
1.1 noro 2899: prev_nm_free_list = _nm_free_list;
2900: prev_ndp_free_list = _ndp_free_list;
2901: _nm_free_list = 0;
2902: _ndp_free_list = 0;
1.53 noro 2903: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2904: if ( trace )
2905: for ( i = nd_psn-1; i >= 0; i-- )
2906: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2907: s0 = 0;
2908: for ( t = d; t; t = NEXT(t) ) {
2909: NEXTND_pairs(s0,s);
2910: s->i1 = t->i1;
2911: s->i2 = t->i2;
1.14 noro 2912: SG(s) = SG(t);
1.61 noro 2913: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2914: }
1.37 noro 2915:
2916: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2917: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2918: old_red[i] = nd_red[i];
2919: nd_red[i] = 0;
2920: }
2921: for ( i = 0; i < REDTAB_LEN; i++ )
2922: for ( r = old_red[i]; r; r = NEXT(r) ) {
2923: NEWRHist(mr);
1.13 noro 2924: mr->index = r->index;
1.20 noro 2925: SG(mr) = SG(r);
1.61 noro 2926: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2927: h = ndl_hash_value(DL(mr));
2928: NEXT(mr) = nd_red[h];
2929: nd_red[h] = mr;
1.6 noro 2930: }
1.37 noro 2931: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2932: old_red = 0;
1.11 noro 2933: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2934: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2935: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2936: nd_psh[i] = r;
1.11 noro 2937: }
1.1 noro 2938: if ( s0 ) NEXT(s) = 0;
2939: prev_nm_free_list = 0;
2940: prev_ndp_free_list = 0;
1.71 noro 2941: #if 0
1.1 noro 2942: GC_gcollect();
1.71 noro 2943: #endif
1.1 noro 2944: return s0;
2945: }
2946:
1.61 noro 2947: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2948: {
1.57 noro 2949: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2950: struct order_pair *op;
1.1 noro 2951:
2952: n = nd_nvar;
1.61 noro 2953: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2954: omask0 = (1<<obpe)-1;
1.34 noro 2955: TD(r) = TD(d);
1.41 noro 2956: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2957: if ( nd_blockmask ) {
2958: l = nd_blockmask->n;
2959: op = nd_blockmask->order_pair;
2960: for ( i = 1; i < nd_exporigin; i++ )
2961: r[i] = d[i];
2962: for ( j = 0, s = 0; j < l; j++ ) {
2963: ord_l = op[j].length;
1.57 noro 2964: for ( i = 0; i < ord_l; i++, s++ ) {
2965: ei = GET_EXP_OLD(d,s);
2966: PUT_EXP(r,s,ei);
2967: }
1.1 noro 2968: }
1.43 noro 2969: } else {
1.56 noro 2970: for ( i = 0; i < n; i++ ) {
2971: ei = GET_EXP_OLD(d,i);
2972: PUT_EXP(r,i,ei);
2973: }
1.1 noro 2974: }
2975: }
1.3 noro 2976:
1.6 noro 2977: ND nd_copy(ND p)
2978: {
2979: NM m,mr,mr0;
1.41 noro 2980: int c,n;
1.6 noro 2981: ND r;
2982:
2983: if ( !p )
2984: return 0;
2985: else {
2986: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2987: NEXTNM(mr0,mr);
1.14 noro 2988: CM(mr) = CM(m);
2989: ndl_copy(DL(m),DL(mr));
1.6 noro 2990: }
2991: NEXT(mr) = 0;
1.31 noro 2992: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2993: SG(r) = SG(p);
1.6 noro 2994: return r;
2995: }
2996: }
2997:
1.53 noro 2998: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2999: {
3000: NM m;
3001: NDV p1,p2;
3002: ND t1,t2;
1.61 noro 3003: UINT *lcm;
1.31 noro 3004: int td;
1.11 noro 3005:
1.77 ! noro 3006: if ( !mod && nd_demand ) {
! 3007: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3008: } else {
1.77 ! noro 3009: if ( trace ) {
! 3010: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
! 3011: } else {
! 3012: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
! 3013: }
1.20 noro 3014: }
1.34 noro 3015: lcm = LCM(p);
1.11 noro 3016: NEWNM(m);
1.20 noro 3017: CQ(m) = HCQ(p2);
1.34 noro 3018: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3019: if ( ndl_check_bound2(p->i1,DL(m)) )
3020: return 0;
1.55 noro 3021: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3022: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3023: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 3024: else chsgnq(HCQ(p1),&CQ(m));
3025: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3026: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3027: nd_free(t1);
3028: return 0;
3029: }
1.55 noro 3030: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3031: *rp = nd_add(mod,t1,t2);
1.11 noro 3032: FREENM(m);
3033: return 1;
3034: }
3035:
1.19 noro 3036: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3037: {
3038: NMV m;
3039: int c,c1,len,i;
3040:
1.34 noro 3041: if ( !p ) return;
1.14 noro 3042: len = LEN(p);
1.71 noro 3043: if ( mod == -1 )
3044: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3045: CM(m) = _mulsf(CM(m),mul);
3046: else
3047: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3048: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3049: }
1.11 noro 3050: }
3051:
1.16 noro 3052: void ndv_mul_c_q(NDV p,Q mul)
3053: {
3054: NMV m;
3055: Q c;
3056: int len,i;
3057:
1.34 noro 3058: if ( !p ) return;
1.16 noro 3059: len = LEN(p);
3060: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3061: mulq(CQ(m),mul,&c); CQ(m) = c;
3062: }
3063: }
3064:
1.55 noro 3065: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3066: int n2,i,j,l,n,tlen;
1.61 noro 3067: UINT *d0;
1.55 noro 3068: NM *tab,*psum;
3069: ND s,r;
3070: NM t;
3071: NMV m1;
3072:
3073: if ( !p ) return 0;
3074: n = NV(p); n2 = n>>1;
3075: d0 = DL(m0);
3076: l = LEN(p);
3077: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3078: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3079: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3080: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3081: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3082: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3083: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3084: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3085: for ( j = 0; j < tlen; j++ ) {
3086: if ( tab[j] ) {
3087: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3088: }
3089: }
3090: }
3091: for ( i = tlen-1, r = 0; i >= 0; i-- )
3092: if ( psum[i] ) {
3093: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3094: MKND(n,psum[i],j,s);
3095: r = nd_add(mod,r,s);
3096: }
1.56 noro 3097: if ( r ) SG(r) = SG(p)+TD(d0);
3098: return r;
1.55 noro 3099: }
3100:
1.56 noro 3101: /* product of monomials */
3102: /* XXX block order is not handled correctly */
3103:
1.55 noro 3104: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3105: {
1.56 noro 3106: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3107: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3108: Q *ctab_q;
3109: Q q,q1;
1.61 noro 3110: UINT c0,c1,c;
1.55 noro 3111: NM *p;
3112: NM m,t;
3113:
3114: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3115: if ( !m0 || !m1 ) return;
3116: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3117: NEWNM(m); d = DL(m);
1.56 noro 3118: if ( mod ) {
3119: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3120: } else
3121: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3122: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3123: homo = n&1 ? 1 : 0;
3124: if ( homo ) {
3125: /* offset of h-degree */
3126: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3127: PUT_EXP(DL(m),n-1,h);
3128: TD(DL(m)) = h;
1.61 noro 3129: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3130: }
3131: tab[0] = m;
3132: NEWNM(m); d = DL(m);
1.57 noro 3133: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3134: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3135: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3136: /* xi^a*(Di^k*xi^l)*Di^b */
3137: a += l; b += k;
1.56 noro 3138: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3139: if ( !k || !l ) {
3140: for ( j = 0; j < curlen; j++ )
1.56 noro 3141: if ( t = tab[j] ) {
3142: dt = DL(t);
3143: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3144: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3145: }
3146: curlen *= k+1;
3147: continue;
3148: }
3149: min = MIN(k,l);
1.56 noro 3150: if ( mod ) {
1.61 noro 3151: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3152: mkwcm(k,l,mod,ctab);
3153: } else {
3154: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3155: mkwc(k,l,ctab_q);
3156: }
1.57 noro 3157: for ( j = min; j >= 0; j-- ) {
1.56 noro 3158: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3159: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3160: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3161: if ( homo ) {
3162: TD(d) = s;
1.56 noro 3163: PUT_EXP(d,n-1,s-h);
1.55 noro 3164: } else TD(d) = h;
1.61 noro 3165: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3166: if ( mod ) c = ctab[j];
3167: else q = ctab_q[j];
1.57 noro 3168: p = tab+curlen*j;
3169: if ( j == 0 ) {
3170: for ( u = 0; u < curlen; u++, p++ ) {
3171: if ( tab[u] ) {
3172: ndl_addto(DL(tab[u]),d);
3173: if ( mod ) {
3174: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3175: } else {
3176: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3177: }
3178: }
1.56 noro 3179: }
1.57 noro 3180: } else {
3181: for ( u = 0; u < curlen; u++, p++ ) {
3182: if ( tab[u] ) {
3183: NEWNM(t);
3184: ndl_add(DL(tab[u]),d,DL(t));
3185: if ( mod ) {
3186: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3187: } else
3188: mulq(CQ(tab[u]),q,&CQ(t));
3189: *p = t;
3190: }
1.55 noro 3191: }
3192: }
3193: }
3194: curlen *= k+1;
3195: }
3196: FREENM(m);
3197: }
3198:
1.63 noro 3199: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3200: {
3201: NM mr,mr0;
3202: NMV m;
3203: UINT *d,*dt,*dm;
3204: int c,n,td,i,c1,c2,len;
3205: Q q;
3206: ND r;
3207:
3208: if ( !p ) return 0;
3209: else {
3210: n = NV(p); m = BDY(p);
3211: d = DL(m0);
3212: len = LEN(p);
3213: mr0 = 0;
3214: td = TD(d);
3215: c = CM(m0);
3216: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3217: NEXTNM(mr0,mr);
3218: CM(mr) = 1;
3219: ndl_add(DL(m),d,DL(mr));
3220: }
3221: NEXT(mr) = 0;
3222: MKND(NV(p),mr0,len,r);
3223: SG(r) = SG(p) + TD(d);
3224: return r;
3225: }
3226: }
3227:
1.55 noro 3228: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3229: {
3230: NM mr,mr0;
3231: NMV m;
1.61 noro 3232: UINT *d,*dt,*dm;
1.9 noro 3233: int c,n,td,i,c1,c2,len;
1.16 noro 3234: Q q;
1.9 noro 3235: ND r;
3236:
1.34 noro 3237: if ( !p ) return 0;
1.55 noro 3238: else if ( do_weyl )
1.71 noro 3239: if ( mod == -1 )
3240: error("ndv_mul_nm : not implemented (weyl)");
3241: else
3242: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3243: else {
3244: n = NV(p); m = BDY(p);
1.34 noro 3245: d = DL(m0);
1.14 noro 3246: len = LEN(p);
1.9 noro 3247: mr0 = 0;
1.34 noro 3248: td = TD(d);
1.71 noro 3249: if ( mod == -1 ) {
3250: c = CM(m0);
3251: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3252: NEXTNM(mr0,mr);
3253: CM(mr) = _mulsf(CM(m),c);
3254: ndl_add(DL(m),d,DL(mr));
3255: }
3256: } else if ( mod ) {
1.16 noro 3257: c = CM(m0);
3258: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3259: NEXTNM(mr0,mr);
3260: c1 = CM(m);
1.19 noro 3261: DMAR(c1,c,0,mod,c2);
1.16 noro 3262: CM(mr) = c2;
3263: ndl_add(DL(m),d,DL(mr));
3264: }
3265: } else {
3266: q = CQ(m0);
3267: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3268: NEXTNM(mr0,mr);
3269: mulq(CQ(m),q,&CQ(mr));
3270: ndl_add(DL(m),d,DL(mr));
3271: }
1.4 noro 3272: }
1.9 noro 3273: NEXT(mr) = 0;
1.31 noro 3274: MKND(NV(p),mr0,len,r);
1.34 noro 3275: SG(r) = SG(p) + TD(d);
1.9 noro 3276: return r;
1.4 noro 3277: }
3278: }
3279:
1.43 noro 3280: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3281: {
1.13 noro 3282: NMV m,mr,mr0,t;
3283: int len,i,k;
1.11 noro 3284:
1.61 noro 3285: if ( !p ) return;
3286: m = BDY(p); len = LEN(p);
3287: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3288: m = (NMV)((char *)mr0+(len-1)*oadv);
3289: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3290: t = (NMV)ALLOCA(nmv_adv);
3291: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3292: CQ(t) = CQ(m);
3293: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3294: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3295: CQ(mr) = CQ(t);
3296: ndl_copy(DL(t),DL(mr));
3297: }
3298: BDY(p) = mr0;
3299: }
3300:
3301: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3302: {
3303: NMV m,mr,mr0;
3304: int len,i;
3305: NDV r;
1.11 noro 3306:
1.61 noro 3307: if ( !p ) return 0;
3308: m = BDY(p); len = LEN(p);
3309: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3310: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3311: ndl_zero(DL(mr));
3312: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3313: CQ(mr) = CQ(m);
1.11 noro 3314: }
1.61 noro 3315: MKNDV(NV(p),mr0,len,r);
3316: SG(r) = SG(p);
3317: return r;
1.11 noro 3318: }
3319:
1.61 noro 3320: /* duplicate p */
3321:
3322: NDV ndv_dup(int mod,NDV p)
1.3 noro 3323: {
3324: NDV d;
1.61 noro 3325: NMV t,m,m0;
1.3 noro 3326: int i,len;
3327:
1.34 noro 3328: if ( !p ) return 0;
1.31 noro 3329: len = LEN(p);
1.34 noro 3330: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3331: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3332: ndl_copy(DL(t),DL(m));
1.16 noro 3333: CQ(m) = CQ(t);
1.3 noro 3334: }
3335: MKNDV(NV(p),m0,len,d);
1.23 noro 3336: SG(d) = SG(p);
3337: return d;
3338: }
3339:
1.63 noro 3340: ND nd_dup(ND p)
3341: {
3342: ND d;
3343: NM t,m,m0;
3344:
3345: if ( !p ) return 0;
3346: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3347: NEXTNM(m0,m);
3348: ndl_copy(DL(t),DL(m));
3349: CQ(m) = CQ(t);
3350: }
3351: if ( m0 ) NEXT(m) = 0;
3352: MKND(NV(p),m0,LEN(p),d);
3353: SG(d) = SG(p);
3354: return d;
3355: }
3356:
1.61 noro 3357: /* XXX if p->len == 0 then it represents 0 */
3358:
3359: void ndv_mod(int mod,NDV p)
3360: {
3361: NMV t,d;
3362: int r;
3363: int i,len,dlen;
1.71 noro 3364: Obj gfs;
1.61 noro 3365:
3366: if ( !p ) return;
3367: len = LEN(p);
3368: dlen = 0;
1.71 noro 3369: if ( mod == -1 )
3370: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3371: simp_ff((Obj)CP(t),&gfs);
3372: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3373: CM(d) = r;
3374: ndl_copy(DL(t),DL(d));
3375: NMV_ADV(d);
3376: dlen++;
3377: }
1.71 noro 3378: else
3379: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3380: r = rem(NM(CQ(t)),mod);
3381: if ( r ) {
3382: if ( SGN(CQ(t)) < 0 )
3383: r = mod-r;
3384: CM(d) = r;
3385: ndl_copy(DL(t),DL(d));
3386: NMV_ADV(d);
3387: dlen++;
3388: }
3389: }
1.61 noro 3390: LEN(p) = dlen;
3391: }
3392:
3393: NDV ptondv(VL vl,VL dvl,P p)
3394: {
3395: ND nd;
3396:
3397: nd = ptond(vl,dvl,p);
3398: return ndtondv(0,nd);
3399: }
3400:
3401: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3402: {
1.61 noro 3403: int n,i,j,k,e;
3404: VL tvl;
3405: V v;
3406: DCP dc;
3407: DCP *w;
3408: ND r,s,t,u;
3409: P x;
3410: int c;
3411: UINT *d;
1.23 noro 3412: NM m,m0;
1.61 noro 3413:
3414: if ( !p )
3415: return 0;
3416: else if ( NUM(p) ) {
3417: NEWNM(m);
3418: ndl_zero(DL(m));
3419: CQ(m) = (Q)p;
3420: NEXT(m) = 0;
3421: MKND(nd_nvar,m,1,r);
3422: SG(r) = 0;
3423: return r;
3424: } else {
3425: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3426: w = (DCP *)ALLOCA(k*sizeof(DCP));
3427: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3428: for ( i = 0, tvl = dvl, v = VR(p);
3429: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3430: if ( !tvl ) {
3431: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3432: t = ptond(vl,dvl,COEF(w[j]));
3433: pwrp(vl,x,DEG(w[j]),&p);
3434: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3435: }
3436: return s;
3437: } else {
3438: NEWNM(m0); d = DL(m0);
3439: for ( j = k-1, s = 0; j >= 0; j-- ) {
3440: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3441: TD(d) = MUL_WEIGHT(e,i);
3442: if ( nd_blockmask) ndl_weight_mask(d);
3443: t = ptond(vl,dvl,COEF(w[j]));
3444: for ( m = BDY(t); m; m = NEXT(m) )
3445: ndl_addto(DL(m),d);
3446: SG(t) += TD(d);
3447: s = nd_add(0,s,t);
3448: }
3449: FREENM(m0);
3450: return s;
3451: }
3452: }
3453: }
3454:
3455: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3456: {
3457: VL tvl;
3458: int len,n,j,i,e;
3459: NMV m;
3460: Q q;
3461: P c;
3462: UINT *d;
3463: P s,r,u,t,w;
1.71 noro 3464: GFS gfs;
1.23 noro 3465:
1.34 noro 3466: if ( !p ) return 0;
1.61 noro 3467: else {
3468: len = LEN(p);
3469: n = NV(p);
3470: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3471: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3472: if ( mod == -1 ) {
3473: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3474: } else if ( mod ) {
1.61 noro 3475: STOQ(CM(m),q); c = (P)q;
3476: } else
3477: c = CP(m);
3478: d = DL(m);
3479: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3480: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3481: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3482: }
3483: addp(vl,s,t,&u); s = u;
3484: }
3485: return s;
1.23 noro 3486: }
1.3 noro 3487: }
3488:
1.61 noro 3489: NDV ndtondv(int mod,ND p)
1.11 noro 3490: {
3491: NDV d;
1.61 noro 3492: NMV m,m0;
3493: NM t;
3494: int i,len;
1.11 noro 3495:
1.34 noro 3496: if ( !p ) return 0;
1.61 noro 3497: len = LEN(p);
1.74 noro 3498: if ( mod )
3499: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3500: else
3501: m0 = m = MALLOC(len*nmv_adv);
3502: ndv_alloc += nmv_adv*len;
1.61 noro 3503: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3504: ndl_copy(DL(t),DL(m));
3505: CQ(m) = CQ(t);
1.11 noro 3506: }
1.61 noro 3507: MKNDV(NV(p),m0,len,d);
1.14 noro 3508: SG(d) = SG(p);
1.11 noro 3509: return d;
3510: }
3511:
1.61 noro 3512: ND ndvtond(int mod,NDV p)
1.11 noro 3513: {
1.61 noro 3514: ND d;
3515: NM m,m0;
1.11 noro 3516: NMV t;
1.61 noro 3517: int i,len;
1.11 noro 3518:
1.34 noro 3519: if ( !p ) return 0;
1.11 noro 3520: m0 = 0;
1.61 noro 3521: len = p->len;
3522: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3523: NEXTNM(m0,m);
3524: ndl_copy(DL(t),DL(m));
3525: CQ(m) = CQ(t);
1.11 noro 3526: }
3527: NEXT(m) = 0;
1.61 noro 3528: MKND(NV(p),m0,len,d);
1.14 noro 3529: SG(d) = SG(p);
1.11 noro 3530: return d;
3531: }
3532:
1.3 noro 3533: void ndv_print(NDV p)
3534: {
3535: NMV m;
3536: int i,len;
3537:
1.34 noro 3538: if ( !p ) printf("0\n");
1.3 noro 3539: else {
1.14 noro 3540: len = LEN(p);
1.3 noro 3541: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3542: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3543: else printf("+%d*",CM(m));
1.16 noro 3544: ndl_print(DL(m));
3545: }
3546: printf("\n");
3547: }
3548: }
3549:
3550: void ndv_print_q(NDV p)
3551: {
3552: NMV m;
3553: int i,len;
3554:
1.34 noro 3555: if ( !p ) printf("0\n");
1.16 noro 3556: else {
3557: len = LEN(p);
3558: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3559: printf("+");
1.74 noro 3560: printexpr(CO,(Obj)CQ(m));
1.16 noro 3561: printf("*");
1.14 noro 3562: ndl_print(DL(m));
1.3 noro 3563: }
3564: printf("\n");
3565: }
1.25 noro 3566: }
3567:
1.61 noro 3568: NODE ndv_reducebase(NODE x)
1.27 noro 3569: {
3570: int len,i,j;
3571: NDV *w;
3572: NODE t,t0;
3573:
3574: len = length(x);
3575: w = (NDV *)ALLOCA(len*sizeof(NDV));
3576: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3577: for ( i = 0; i < len; i++ ) {
3578: for ( j = 0; j < i; j++ ) {
3579: if ( w[i] && w[j] )
3580: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3581: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3582: }
3583: }
3584: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3585: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3586: }
3587: NEXT(t) = 0; x = t0;
3588: return x;
1.11 noro 3589: }
1.32 noro 3590:
1.43 noro 3591: /* XXX incomplete */
3592:
1.32 noro 3593: void nd_init_ord(struct order_spec *ord)
3594: {
1.43 noro 3595: switch ( ord->id ) {
1.32 noro 3596: case 0:
1.43 noro 3597: switch ( ord->ord.simple ) {
3598: case 0:
3599: nd_dcomp = 1;
3600: nd_isrlex = 1;
3601: break;
3602: case 1:
3603: nd_dcomp = 1;
3604: nd_isrlex = 0;
3605: break;
3606: case 2:
3607: nd_dcomp = 0;
3608: nd_isrlex = 0;
1.45 noro 3609: ndl_compare_function = ndl_lex_compare;
1.58 noro 3610: break;
3611: case 11:
3612: /* XXX */
3613: nd_dcomp = 0;
3614: nd_isrlex = 1;
3615: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3616: break;
3617: default:
3618: error("nd_gr : unsupported order");
3619: }
1.32 noro 3620: break;
3621: case 1:
1.43 noro 3622: /* XXX */
3623: nd_dcomp = -1;
1.32 noro 3624: nd_isrlex = 0;
1.45 noro 3625: ndl_compare_function = ndl_block_compare;
1.34 noro 3626: break;
1.43 noro 3627: case 2:
3628: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3629: break;
3630: }
1.41 noro 3631: nd_ord = ord;
1.32 noro 3632: }
3633:
1.43 noro 3634: BlockMask nd_create_blockmask(struct order_spec *ord)
3635: {
3636: int n,i,j,s,l;
1.61 noro 3637: UINT *t;
1.43 noro 3638: BlockMask bm;
3639:
3640: if ( !ord->id )
3641: return 0;
3642: n = ord->ord.block.length;
3643: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3644: bm->n = n;
3645: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3646: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3647: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3648: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3649: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3650: l = bm->order_pair[i].length;
3651: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3652: }
3653: return bm;
1.57 noro 3654: }
3655:
3656: EPOS nd_create_epos(struct order_spec *ord)
3657: {
3658: int i,j,l,s,ord_l,ord_o;
3659: EPOS epos;
3660: struct order_pair *op;
3661:
3662: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3663: switch ( ord->id ) {
3664: case 0:
3665: if ( nd_isrlex ) {
3666: for ( i = 0; i < nd_nvar; i++ ) {
3667: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3668: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3669: }
3670: } else {
3671: for ( i = 0; i < nd_nvar; i++ ) {
3672: epos[i].i = nd_exporigin + i/nd_epw;
3673: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3674: }
3675: }
3676: break;
3677: case 1:
3678: /* block order */
3679: l = ord->ord.block.length;
3680: op = ord->ord.block.order_pair;
3681: for ( j = 0, s = 0; j < l; j++ ) {
3682: ord_o = op[j].order;
3683: ord_l = op[j].length;
3684: if ( !ord_o )
3685: for ( i = 0; i < ord_l; i++ ) {
3686: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3687: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3688: }
3689: else
3690: for ( i = 0; i < ord_l; i++ ) {
3691: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3692: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3693: }
3694: s += ord_l;
3695: }
3696: break;
3697: case 2:
3698: error("nd_create_epos : matrix order is not supported yet.");
3699: }
3700: return epos;
1.43 noro 3701: }
1.59 noro 3702:
3703: /* external interface */
3704:
3705: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3706: {
1.61 noro 3707: NODE t,in0,in;
1.59 noro 3708: ND nd,nf;
1.61 noro 3709: NDV ndv;
3710: VL vv,tv;
3711: int stat,nvar,max,e;
1.69 noro 3712: union oNDC dn;
1.59 noro 3713:
3714: pltovl(v,&vv);
1.74 noro 3715: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3716:
3717: /* get the degree bound */
3718: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3719: for ( tv = vv; tv; tv = NEXT(tv) ) {
3720: e = getdeg(tv->v,(P)BDY(t));
3721: max = MAX(e,max);
3722: }
3723: for ( tv = vv; tv; tv = NEXT(tv) ) {
3724: e = getdeg(tv->v,f);
3725: max = MAX(e,max);
3726: }
3727:
1.59 noro 3728: nd_init_ord(ord);
1.61 noro 3729: nd_setup_parameters(nvar,max);
3730:
3731: /* conversion to ndv */
3732: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3733: NEXTNODE(in0,in);
3734: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3735: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3736: }
3737: NEXTNODE(in0,in);
3738: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3739: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3740: NEXT(in) = 0;
3741:
3742: ndv_setup(m,0,in0);
1.59 noro 3743: nd_psn--;
3744: nd_scale=2;
3745: while ( 1 ) {
3746: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3747: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3748: if ( !stat ) {
3749: nd_psn++;
3750: nd_reconstruct(m,0,0);
3751: nd_psn--;
3752: } else
3753: break;
3754: }
1.61 noro 3755: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3756: }
3757:
3758: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3759: {
3760: NM m;
3761: UINT *t,*s;
3762: int i;
3763:
3764: for ( i = 0; i < n; i++ ) r[i] = 0;
3765: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3766: t = DL(m);
3767: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3768: r[i] = CM(m);
3769: }
3770: for ( i = 0; !r[i]; i++ );
3771: return i;
3772: }
3773:
1.74 noro 3774: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3775: {
3776: NMV m;
3777: UINT *t,*s;
3778: int i,j,len;
3779:
3780: for ( i = 0; i < n; i++ ) r[i] = 0;
3781: m = BDY(d);
3782: len = LEN(d);
3783: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3784: t = DL(m);
3785: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3786: r[i] = CM(m);
3787: }
3788: for ( i = 0; !r[i]; i++ );
3789: return i;
3790: }
3791:
1.63 noro 3792: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3793: {
3794: NM m;
3795: NMV mr;
3796: UINT *d,*t,*s;
3797: NDV p;
3798: int i,j,len;
3799:
3800: m = pair->mul;
3801: d = DL(m);
3802: p = nd_ps[pair->index];
3803: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3804: for ( i = 0; i < n; i++ ) r[i] = 0;
3805: len = LEN(p);
3806: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3807: ndl_add(d,DL(mr),t);
3808: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3809: r[i] = CM(mr);
3810: }
3811: for ( i = 0; !r[i]; i++ );
3812: return i;
3813: }
3814:
1.67 noro 3815: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3816: {
3817: NM m;
3818: NMV mr;
3819: UINT *d,*t,*s;
3820: NDV p;
1.67 noro 3821: unsigned char *ivc;
3822: unsigned short *ivs;
3823: UINT *v,*ivi;
3824: int i,j,len,prev,diff,cdiff;
3825: IndArray r;
1.64 noro 3826:
3827: m = pair->mul;
3828: d = DL(m);
3829: p = nd_ps[pair->index];
3830: len = LEN(p);
3831: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3832: r = (IndArray)MALLOC(sizeof(struct oIndArray));
3833: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3834: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3835: ndl_add(d,DL(mr),t);
3836: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3837: v[j] = i;
3838: }
3839: r->head = v[0];
3840: diff = 0;
3841: for ( i = 1; i < len; i++ ) {
3842: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3843: }
3844: if ( diff < 256 ) {
3845: r->width = 1;
3846: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3847: r->index.c = ivc;
3848: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3849: } else if ( diff < 65536 ) {
3850: r->width = 2;
3851: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3852: r->index.s = ivs;
3853: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3854: } else {
3855: r->width = 4;
3856: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3857: r->index.i = ivi;
3858: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3859: }
1.67 noro 3860: return r;
1.64 noro 3861: }
3862:
1.65 noro 3863:
1.76 noro 3864: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3865: {
1.67 noro 3866: int i,j,k,len,pos,prev;
1.66 noro 3867: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3868: IndArray ivect;
3869: unsigned char *ivc;
3870: unsigned short *ivs;
3871: unsigned int *ivi;
1.65 noro 3872: NDV redv;
1.67 noro 3873: NMV mr;
1.65 noro 3874: NODE rp;
1.76 noro 3875: int maxrs;
1.65 noro 3876:
1.76 noro 3877: maxrs = 0;
1.74 noro 3878: for ( i = 0; i < nred; i++ ) {
1.65 noro 3879: ivect = imat[i];
1.67 noro 3880: k = ivect->head; svect[k] %= m;
1.65 noro 3881: if ( c = svect[k] ) {
1.76 noro 3882: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3883: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3884: len = LEN(redv); mr = BDY(redv);
3885: svect[k] = 0; prev = k;
3886: switch ( ivect->width ) {
3887: case 1:
3888: ivc = ivect->index.c;
3889: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3890: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3891: prev = pos;
3892: DMA(c1,c,c2,up,lo);
3893: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3894: } else svect[pos] = lo;
3895: }
3896: break;
3897: case 2:
3898: ivs = ivect->index.s;
3899: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3900: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3901: prev = pos;
3902: DMA(c1,c,c2,up,lo);
3903: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3904: } else svect[pos] = lo;
3905: }
3906: break;
3907: case 4:
3908: ivi = ivect->index.i;
3909: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3910: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3911: prev = pos;
3912: DMA(c1,c,c2,up,lo);
3913: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3914: } else svect[pos] = lo;
3915: }
3916: break;
1.65 noro 3917: }
3918: }
3919: }
1.66 noro 3920: for ( i = 0; i < col; i++ )
3921: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3922: return maxrs;
1.65 noro 3923: }
3924:
1.76 noro 3925: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3926: {
3927: int i,j,k,len,pos,prev;
3928: UINT c,c1,c2,c3,up,lo,dmy;
3929: IndArray ivect;
3930: unsigned char *ivc;
3931: unsigned short *ivs;
3932: unsigned int *ivi;
3933: NDV redv;
3934: NMV mr;
3935: NODE rp;
1.76 noro 3936: int maxrs;
1.72 noro 3937:
1.76 noro 3938: maxrs = 0;
1.74 noro 3939: for ( i = 0; i < nred; i++ ) {
1.72 noro 3940: ivect = imat[i];
3941: k = ivect->head; svect[k] %= m;
3942: if ( c = svect[k] ) {
1.76 noro 3943: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3944: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3945: len = LEN(redv); mr = BDY(redv);
3946: svect[k] = 0; prev = k;
3947: switch ( ivect->width ) {
3948: case 1:
3949: ivc = ivect->index.c;
3950: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3951: pos = prev+ivc[j]; prev = pos;
3952: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3953: }
3954: break;
3955: case 2:
3956: ivs = ivect->index.s;
3957: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3958: pos = prev+ivs[j]; prev = pos;
3959: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3960: }
3961: break;
3962: case 4:
3963: ivi = ivect->index.i;
3964: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3965: pos = prev+ivi[j]; prev = pos;
3966: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3967: }
3968: break;
3969: }
3970: }
3971: }
1.76 noro 3972: return maxrs;
1.72 noro 3973: }
3974:
1.65 noro 3975: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3976: {
3977: int j,k,len;
3978: UINT *p;
3979: UINT c;
3980: NDV r;
3981: NMV mr0,mr;
3982:
3983: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3984: if ( !len ) return 0;
3985: else {
1.74 noro 3986: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
3987: ndv_alloc += nmv_adv*len;
1.65 noro 3988: mr = mr0;
3989: p = s0vect;
3990: for ( j = k = 0; j < col; j++, p += nd_wpd )
3991: if ( !rhead[j] ) {
3992: if ( c = vect[k++] ) {
3993: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3994: }
3995: }
3996: MKNDV(nd_nvar,mr0,len,r);
3997: return r;
3998: }
3999: }
4000:
1.74 noro 4001: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4002: {
4003: ND_pairs t;
4004: NODE sp0,sp;
4005: int stat;
4006: ND spol;
4007:
4008: for ( t = l; t; t = NEXT(t) ) {
4009: stat = nd_sp(m,0,t,&spol);
4010: if ( !stat ) return 0;
4011: if ( spol ) {
4012: add_pbucket_symbolic(bucket,spol);
4013: }
4014: }
1.68 noro 4015: return 1;
1.65 noro 4016: }
4017:
4018: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4019: {
4020: NODE rp0,rp;
4021: NM mul,head,s0,s;
1.76 noro 4022: int index,col,i,sugar;
1.65 noro 4023: RHist h;
4024: UINT *s0v,*p;
4025: NM_ind_pair pair;
4026: ND red;
4027:
4028: s0 = 0; rp0 = 0; col = 0;
4029: while ( 1 ) {
4030: head = remove_head_pbucket_symbolic(bucket);
4031: if ( !head ) break;
4032: if ( !s0 ) s0 = head;
4033: else NEXT(s) = head;
4034: s = head;
4035: index = ndl_find_reducer(DL(head));
4036: if ( index >= 0 ) {
4037: h = nd_psh[index];
4038: NEWNM(mul);
4039: ndl_sub(DL(head),DL(h),DL(mul));
4040: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4041: sugar = TD(DL(mul))+SG(nd_ps[index]);
4042: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4043: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4044: add_pbucket_symbolic(bucket,nd_remove_head(red));
4045: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4046: }
4047: col++;
4048: }
1.72 noro 4049: if ( rp0 ) NEXT(rp) = 0;
4050: NEXT(s) = 0;
1.65 noro 4051: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4052: for ( i = 0, p = s0v, s = s0; i < col;
4053: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4054: *s0vect = s0v;
4055: *r = rp0;
4056: return col;
4057: }
4058:
1.69 noro 4059: NODE nd_f4(int m)
4060: {
4061: int i,nh,stat,index;
4062: NODE r,g;
4063: ND_pairs d,l,t;
4064: ND spol,red;
4065: NDV nf,redv;
4066: NM s0,s;
1.74 noro 4067: NODE rp0,sp0,srp0,nflist;
1.69 noro 4068: int nsp,nred,col,rank,len,k,j,a;
4069: UINT c;
1.74 noro 4070: UINT **spmat;
1.69 noro 4071: UINT *s0vect,*svect,*p,*v;
4072: int *colstat;
4073: IndArray *imat;
4074: int *rhead;
4075: int spcol,sprow;
4076: int sugar;
4077: PGeoBucket bucket;
4078: struct oEGT eg0,eg1,eg_f4;
4079:
4080: if ( !m )
4081: error("nd_f4 : not implemented");
1.74 noro 4082: ndv_alloc = 0;
1.69 noro 4083: g = 0; d = 0;
4084: for ( i = 0; i < nd_psn; i++ ) {
4085: d = update_pairs(d,g,i);
4086: g = update_base(g,i);
4087: }
4088: while ( d ) {
4089: get_eg(&eg0);
4090: l = nd_minsugarp(d,&d);
4091: sugar = SG(l);
4092: bucket = create_pbucket();
1.74 noro 4093: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4094: if ( !stat ) {
4095: for ( t = l; NEXT(t); t = NEXT(t) );
4096: NEXT(t) = d; d = l;
4097: d = nd_reconstruct(m,0,d);
4098: continue;
4099: }
4100: if ( !sp0 ) continue;
4101: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4102: if ( !col ) {
4103: for ( t = l; NEXT(t); t = NEXT(t) );
4104: NEXT(t) = d; d = l;
4105: d = nd_reconstruct(m,0,d);
4106: continue;
4107: }
4108: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4109: if ( DP_Print )
4110: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4111: sugar,eg_f4.exectime+eg_f4.gctime);
4112: if ( 1 )
4113: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4114: else
4115: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4116: /* adding new bases */
1.74 noro 4117: for ( r = nflist; r; r = NEXT(r) ) {
4118: nf = (NDV)BDY(r);
1.69 noro 4119: ndv_removecont(m,nf);
1.77 ! noro 4120: nh = ndv_newps(m,nf,0);
1.69 noro 4121: d = update_pairs(d,g,nh);
4122: g = update_base(g,nh);
4123: }
4124: }
4125: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 4126: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 4127: return g;
4128: }
1.74 noro 4129:
4130: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4131: {
1.67 noro 4132: IndArray *imat;
1.74 noro 4133: int nsp,nred,spcol,sprow,a;
1.65 noro 4134: int *rhead;
1.74 noro 4135: int i,j,k,l,rank;
4136: NODE rp,r0,r;
4137: ND_pairs sp;
4138: ND spol;
4139: int **spmat;
4140: UINT *svect,*v;
4141: int *colstat;
1.66 noro 4142: struct oEGT eg0,eg1,eg_f4;
1.74 noro 4143: NM_ind_pair *rvect;
1.76 noro 4144: int maxrs;
4145: int *spsugar;
1.63 noro 4146:
1.74 noro 4147: get_eg(&eg0);
4148: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4149: nred = length(rp0); spcol = col-nred;
4150: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4151: rhead = (int *)ALLOCA(col*sizeof(int));
4152: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4153:
1.74 noro 4154: /* construction of index arrays */
4155: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4156: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4157: rvect[i] = (NM_ind_pair)BDY(rp);
4158: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4159: rhead[imat[i]->head] = 1;
4160: }
4161:
4162: /* elimination (1st step) */
4163: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4164: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4165: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4166: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4167: nd_sp(m,0,sp,&spol);
1.75 noro 4168: if ( !spol ) continue;
1.74 noro 4169: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4170: if ( m == -1 )
4171: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4172: else
4173: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4174: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4175: if ( i < col ) {
4176: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4177: for ( j = k = 0; j < col; j++ )
4178: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4179: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4180: sprow++;
4181: }
1.76 noro 4182: nd_free(spol);
1.74 noro 4183: }
4184: /* free index arrays */
4185: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4186:
4187: /* elimination (2nd step) */
4188: colstat = (int *)ALLOCA(spcol*sizeof(int));
4189: if ( m == -1 )
1.76 noro 4190: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4191: else
1.76 noro 4192: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4193: r0 = 0;
4194: for ( i = 0; i < rank; i++ ) {
4195: NEXTNODE(r0,r); BDY(r) =
4196: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4197: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4198: GC_free(spmat[i]);
4199: }
4200: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
4201: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4202: if ( DP_Print ) {
4203: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4204: nsp,nred,sprow,spcol,rank);
4205: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4206: }
1.74 noro 4207: return r0;
4208: }
4209:
4210: FILE *nd_write,*nd_read;
4211:
4212: void nd_send_int(int a) {
4213: write_int(nd_write,&a);
4214: }
4215:
4216: void nd_send_intarray(int *p,int len) {
4217: write_intarray(nd_write,p,len);
4218: }
4219:
4220: int nd_recv_int() {
4221: int a;
4222:
4223: read_int(nd_read,&a);
4224: return a;
4225: }
4226:
4227: void nd_recv_intarray(int *p,int len) {
4228: read_intarray(nd_read,p,len);
4229: }
4230:
4231: void nd_send_ndv(NDV p) {
4232: int len,i;
4233: NMV m;
4234:
4235: if ( !p ) nd_send_int(0);
4236: else {
4237: len = LEN(p);
4238: nd_send_int(len);
4239: m = BDY(p);
4240: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4241: nd_send_int(CM(m));
4242: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4243: }
1.74 noro 4244: }
4245: }
4246:
4247: void nd_send_nd(ND p) {
4248: int len,i;
4249: NM m;
4250:
4251: if ( !p ) nd_send_int(0);
4252: else {
4253: len = LEN(p);
4254: nd_send_int(len);
4255: m = BDY(p);
4256: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4257: nd_send_int(CM(m));
4258: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4259: }
1.74 noro 4260: }
4261: }
1.65 noro 4262:
1.74 noro 4263: NDV nd_recv_ndv()
4264: {
4265: int len,i;
4266: NMV m,m0;
4267: NDV r;
1.65 noro 4268:
1.74 noro 4269: len = nd_recv_int();
4270: if ( !len ) return 0;
4271: else {
4272: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4273: ndv_alloc += len*nmv_adv;
4274: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4275: CM(m) = nd_recv_int();
4276: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4277: }
1.74 noro 4278: MKNDV(nd_nvar,m0,len,r);
4279: return r;
4280: }
4281: }
1.65 noro 4282:
1.74 noro 4283: int ox_exec_f4_red(Q proc)
4284: {
4285: Obj obj;
4286: STRING fname;
4287: NODE arg;
4288: int s;
4289: extern int ox_need_conv,ox_file_io;
4290:
4291: MKSTR(fname,"nd_exec_f4_red");
4292: arg = mknode(2,proc,fname);
4293: Pox_cmo_rpc(arg,&obj);
4294: s = get_ox_server_id(QTOS(proc));
4295: nd_write = iofp[s].out;
4296: nd_read = iofp[s].in;
4297: ox_need_conv = ox_file_io = 0;
4298: return s;
4299: }
4300:
4301: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4302: {
4303: int nsp,nred;
4304: int i,rank,s;
4305: NODE rp,r0,r;
4306: ND_pairs sp;
4307: NM_ind_pair pair;
4308: NMV nmv;
4309: NM nm;
4310: NDV nf;
4311: Obj proc,dmy;
4312:
4313: ox_launch_main(0,0,&proc);
4314: s = ox_exec_f4_red((Q)proc);
4315:
4316: nd_send_int(m);
4317: nd_send_int(nd_nvar);
4318: nd_send_int(nd_bpe);
4319: nd_send_int(nd_wpd);
4320: nd_send_int(nmv_adv);
4321:
4322: saveobj(nd_write,dp_current_spec.obj); fflush(nd_write);
4323:
4324: nd_send_int(nd_psn);
4325: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4326:
4327: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4328: nd_send_int(nsp);
4329: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4330: nd_send_int(sp->i1); nd_send_int(sp->i2);
4331: }
4332:
4333: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4334:
4335: nred = length(rp0); nd_send_int(nred);
4336: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4337: pair = (NM_ind_pair)BDY(rp);
4338: nd_send_int(pair->index);
4339: nd_send_intarray(pair->mul->dl,nd_wpd);
4340: }
4341: fflush(nd_write);
4342: rank = nd_recv_int();
4343: fprintf(asir_out,"rank=%d\n",rank);
4344: r0 = 0;
4345: for ( i = 0; i < rank; i++ ) {
4346: nf = nd_recv_ndv();
4347: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4348: }
4349: Pox_shutdown(mknode(1,proc),&dmy);
4350: return r0;
4351: }
4352:
4353: /* server side */
4354:
4355: void nd_exec_f4_red_dist()
4356: {
4357: int m,i,nsp,col,s0size,nred,spcol,j,k;
4358: NM_ind_pair *rp0;
4359: NDV nf;
4360: UINT *s0vect;
4361: IndArray *imat;
4362: int *rhead;
4363: int **spmat;
4364: UINT *svect,*v;
4365: ND_pairs *sp0;
4366: int *colstat;
4367: int a,sprow,rank;
4368: struct order_spec ord;
4369: Obj ordspec;
4370: ND spol;
1.76 noro 4371: int maxrs;
4372: int *spsugar;
1.74 noro 4373:
4374: nd_read = iofp[0].in;
4375: nd_write = iofp[0].out;
4376: m = nd_recv_int();
4377: nd_nvar = nd_recv_int();
4378: nd_bpe = nd_recv_int();
4379: nd_wpd = nd_recv_int();
4380: nmv_adv = nd_recv_int();
4381:
4382: loadobj(nd_read,&ordspec);
4383: create_order_spec(ordspec,&ord);
4384: nd_init_ord(&ord);
4385: nd_setup_parameters(nd_nvar,0);
4386:
4387: nd_psn = nd_recv_int();
4388: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4389: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4390: for ( i = 0; i < nd_psn; i++ ) {
4391: nd_ps[i] = nd_recv_ndv();
4392: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4393: }
4394:
4395: nsp = nd_recv_int();
4396: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4397: for ( i = 0; i < nsp; i++ ) {
4398: NEWND_pairs(sp0[i]);
4399: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4400: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4401: }
4402:
4403: col = nd_recv_int();
4404: s0size = col*nd_wpd;
4405: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4406: nd_recv_intarray(s0vect,s0size);
4407:
4408: nred = nd_recv_int();
4409: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4410: for ( i = 0; i < nred; i++ ) {
4411: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4412: rp0[i]->index = nd_recv_int();
4413: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4414: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4415: }
4416:
4417: spcol = col-nred;
4418: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4419: rhead = (int *)MALLOC(col*sizeof(int));
4420: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4421:
4422: /* construction of index arrays */
4423: for ( i = 0; i < nred; i++ ) {
4424: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4425: rhead[imat[i]->head] = 1;
4426: }
4427:
4428: /* elimination (1st step) */
4429: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4430: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4431: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4432: for ( a = sprow = 0; a < nsp; a++ ) {
4433: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4434: if ( !spol ) continue;
1.74 noro 4435: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4436: if ( m == -1 )
4437: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4438: else
4439: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4440: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4441: if ( i < col ) {
4442: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4443: for ( j = k = 0; j < col; j++ )
4444: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4445: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4446: sprow++;
4447: }
1.76 noro 4448: nd_free(spol);
1.74 noro 4449: }
4450: /* elimination (2nd step) */
4451: colstat = (int *)ALLOCA(spcol*sizeof(int));
4452: if ( m == -1 )
1.76 noro 4453: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4454: else
1.76 noro 4455: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4456: nd_send_int(rank);
4457: for ( i = 0; i < rank; i++ ) {
4458: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4459: nd_send_ndv(nf);
1.63 noro 4460: }
1.74 noro 4461: fflush(nd_write);
1.76 noro 4462: }
4463:
4464: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4465: {
4466: int i,j,k,l,inv,a,rank,s;
4467: unsigned int *t,*pivot,*pk;
4468: unsigned int **mat;
4469:
4470: mat = (unsigned int **)mat0;
4471: for ( rank = 0, j = 0; j < col; j++ ) {
4472: for ( i = rank; i < row; i++ )
4473: mat[i][j] %= md;
4474: for ( i = rank; i < row; i++ )
4475: if ( mat[i][j] )
4476: break;
4477: if ( i == row ) {
4478: colstat[j] = 0;
4479: continue;
4480: } else
4481: colstat[j] = 1;
4482: if ( i != rank ) {
4483: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4484: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4485: }
4486: pivot = mat[rank];
4487: s = sugar[rank];
4488: inv = invm(pivot[j],md);
4489: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4490: if ( *pk ) {
4491: if ( *pk >= (unsigned int)md )
4492: *pk %= md;
4493: DMAR(*pk,inv,0,md,*pk)
4494: }
4495: for ( i = rank+1; i < row; i++ ) {
4496: t = mat[i];
4497: if ( a = t[j] ) {
4498: sugar[i] = MAX(sugar[i],s);
4499: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4500: }
4501: }
4502: rank++;
4503: }
4504: for ( j = col-1, l = rank-1; j >= 0; j-- )
4505: if ( colstat[j] ) {
4506: pivot = mat[l];
4507: s = sugar[l];
4508: for ( i = 0; i < l; i++ ) {
4509: t = mat[i];
4510: t[j] %= md;
4511: if ( a = t[j] ) {
4512: sugar[i] = MAX(sugar[i],s);
4513: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4514: }
4515: }
4516: l--;
4517: }
4518: for ( j = 0, l = 0; l < rank; j++ )
4519: if ( colstat[j] ) {
4520: t = mat[l];
4521: for ( k = j; k < col; k++ )
4522: if ( t[k] >= (unsigned int)md )
4523: t[k] %= md;
4524: l++;
4525: }
4526: return rank;
4527: }
4528:
4529: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4530: {
4531: int i,j,k,l,inv,a,rank,s;
4532: unsigned int *t,*pivot,*pk;
4533: unsigned int **mat;
4534:
4535: mat = (unsigned int **)mat0;
4536: for ( rank = 0, j = 0; j < col; j++ ) {
4537: for ( i = rank; i < row; i++ )
4538: if ( mat[i][j] )
4539: break;
4540: if ( i == row ) {
4541: colstat[j] = 0;
4542: continue;
4543: } else
4544: colstat[j] = 1;
4545: if ( i != rank ) {
4546: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4547: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4548: }
4549: pivot = mat[rank];
4550: s = sugar[rank];
4551: inv = _invsf(pivot[j]);
4552: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4553: if ( *pk )
4554: *pk = _mulsf(*pk,inv);
4555: for ( i = rank+1; i < row; i++ ) {
4556: t = mat[i];
4557: if ( a = t[j] ) {
4558: sugar[i] = MAX(sugar[i],s);
4559: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4560: }
4561: }
4562: rank++;
4563: }
4564: for ( j = col-1, l = rank-1; j >= 0; j-- )
4565: if ( colstat[j] ) {
4566: pivot = mat[l];
4567: s = sugar[l];
4568: for ( i = 0; i < l; i++ ) {
4569: t = mat[i];
4570: if ( a = t[j] ) {
4571: sugar[i] = MAX(sugar[i],s);
4572: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4573: }
4574: }
4575: l--;
4576: }
4577: return rank;
1.77 ! noro 4578: }
! 4579:
! 4580: int ndv_ishomo(NDV p)
! 4581: {
! 4582: NMV m;
! 4583: int len,h;
! 4584:
! 4585: if ( !p ) return 1;
! 4586: len = LEN(p);
! 4587: m = BDY(p);
! 4588: h = TD(DL(m));
! 4589: NMV_ADV(m);
! 4590: for ( len--; len; len--, NMV_ADV(m) )
! 4591: if ( TD(DL(m)) != h ) return 0;
! 4592: return 1;
! 4593: }
! 4594:
! 4595: void ndv_save(NDV p,int index)
! 4596: {
! 4597: FILE *s;
! 4598: char name[BUFSIZ];
! 4599: short id;
! 4600: int nv,sugar,len,n,i,td,e,j;
! 4601: NMV m;
! 4602: unsigned int *dl;
! 4603:
! 4604: sprintf(name,"%s/%d",Demand,index);
! 4605: s = fopen(name,"w");
! 4606: savevl(s,0);
! 4607: if ( !p ) {
! 4608: saveobj(s,0);
! 4609: return;
! 4610: }
! 4611: id = O_DP;
! 4612: nv = NV(p);
! 4613: sugar = SG(p);
! 4614: len = LEN(p);
! 4615: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
! 4616: write_int(s,&len);
! 4617:
! 4618: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
! 4619: saveobj(s,(Obj)CQ(m));
! 4620: dl = DL(m);
! 4621: td = TD(dl);
! 4622: write_int(s,&td);
! 4623: for ( j = 0; j < nv; j++ ) {
! 4624: e = GET_EXP(dl,j);
! 4625: write_int(s,&e);
! 4626: }
! 4627: }
! 4628: fclose(s);
! 4629: }
! 4630:
! 4631: NDV ndv_load(int index)
! 4632: {
! 4633: FILE *s;
! 4634: char name[BUFSIZ];
! 4635: short id;
! 4636: int nv,sugar,len,n,i,td,e,j;
! 4637: NDV d;
! 4638: NMV m0,m;
! 4639: unsigned int *dl;
! 4640: Obj obj;
! 4641:
! 4642: sprintf(name,"%s/%d",Demand,index);
! 4643: s = fopen(name,"r");
! 4644: if ( !s ) return 0;
! 4645:
! 4646: skipvl(s);
! 4647: read_short(s,&id);
! 4648: if ( !id ) return 0;
! 4649: read_int(s,&nv);
! 4650: read_int(s,&sugar);
! 4651: read_int(s,&len);
! 4652:
! 4653: m0 = m = MALLOC(len*nmv_adv);
! 4654: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
! 4655: loadobj(s,&obj); CQ(m) = (Q)obj;
! 4656: dl = DL(m);
! 4657: ndl_zero(dl);
! 4658: read_int(s,&td); TD(dl) = td;
! 4659: for ( j = 0; j < nv; j++ ) {
! 4660: read_int(s,&e);
! 4661: PUT_EXP(dl,j,e);
! 4662: }
! 4663: if ( nd_blockmask ) ndl_weight_mask(dl);
! 4664: }
! 4665: fclose(s);
! 4666: MKNDV(nv,m0,len,d);
! 4667: SG(d) = sugar;
! 4668: return d;
1.59 noro 4669: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>