Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.82
1.82 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.81 2003/10/11 02:14:19 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);
1.82 ! noro 273: NODE nd_gb_trace(int m,int ishomo);
1.40 noro 274:
275: /* ndl functions */
1.61 noro 276: int ndl_weight(UINT *d);
277: void ndl_weight_mask(UINT *d);
278: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight);
279: void ndl_dehomogenize(UINT *p);
280: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos);
281: INLINE int ndl_reducible(UINT *d1,UINT *d2);
282: INLINE int ndl_lex_compare(UINT *d1,UINT *d2);
283: INLINE int ndl_block_compare(UINT *d1,UINT *d2);
284: INLINE int ndl_equal(UINT *d1,UINT *d2);
285: INLINE void ndl_copy(UINT *d1,UINT *d2);
286: INLINE void ndl_zero(UINT *d);
287: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d);
288: INLINE void ndl_addto(UINT *d1,UINT *d2);
289: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d);
290: INLINE int ndl_hash_value(UINT *d);
1.45 noro 291:
292: /* normal forms */
1.63 noro 293: INLINE int ndl_find_reducer(UINT *g);
1.53 noro 294: int nd_sp(int mod,int trace,ND_pairs p,ND *nf);
1.69 noro 295: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *nf);
1.53 noro 296: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *nf);
1.40 noro 297:
298: /* finalizers */
1.61 noro 299: NODE ndv_reducebase(NODE x);
300: NODE ndv_reduceall(int m,NODE f);
1.40 noro 301:
302: /* allocators */
303: void nd_free_private_storage();
304: void _NM_alloc();
305: void _ND_alloc();
1.1 noro 306: void nd_free(ND p);
1.40 noro 307: void nd_free_redlist();
308:
309: /* printing */
1.61 noro 310: void ndl_print(UINT *dl);
1.1 noro 311: void nd_print(ND p);
1.16 noro 312: void nd_print_q(ND p);
1.1 noro 313: void ndp_print(ND_pairs d);
1.40 noro 314:
315:
316: /* setup, reconstruct */
317: void nd_init_ord(struct order_spec *spec);
318: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs ndp);
1.61 noro 319: void ndv_setup(int mod,int trace,NODE f);
320: void nd_setup_parameters(int nvar,int max);
1.43 noro 321: BlockMask nd_create_blockmask(struct order_spec *ord);
1.57 noro 322: EPOS nd_create_epos(struct order_spec *ord);
1.48 noro 323: int nd_get_exporigin(struct order_spec *ord);
1.61 noro 324: void ndv_mod(int mod,NDV p);
325: NDV ndv_dup(int mod,NDV p);
1.63 noro 326: ND nd_dup(ND p);
1.40 noro 327:
328: /* ND functions */
1.61 noro 329: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand);
1.40 noro 330: void nd_mul_c(int mod,ND p,int mul);
331: void nd_mul_c_q(ND p,Q mul);
1.61 noro 332: void nd_mul_c_p(VL vl,ND p,P mul);
1.40 noro 333: ND nd_remove_head(ND p);
1.69 noro 334: ND nd_separate_head(ND p,ND *head);
1.1 noro 335: int nd_length(ND p);
1.61 noro 336: void nd_append_red(UINT *d,int i);
337: UINT *ndv_compute_bound(NDV p);
1.6 noro 338: ND nd_copy(ND p);
1.63 noro 339: ND nd_merge(ND p1,ND p2);
1.40 noro 340: ND nd_add(int mod,ND p1,ND p2);
341: ND nd_add_q(ND p1,ND p2);
1.71 noro 342: ND nd_add_sf(ND p1,ND p2);
1.41 noro 343: INLINE int nd_length(ND p);
1.74 noro 344: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
345: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
1.4 noro 346:
1.40 noro 347: /* NDV functions */
1.55 noro 348: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p);
349: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen);
1.19 noro 350: void ndv_mul_c(int mod,NDV p,int mul);
1.40 noro 351: void ndv_mul_c_q(NDV p,Q mul);
1.63 noro 352: ND ndv_mul_nm_symbolic(NM m0,NDV p);
1.61 noro 353: ND ndv_mul_nm(int mod,NM m0,NDV p);
1.43 noro 354: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos);
1.61 noro 355: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos);
356: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS eops);
1.45 noro 357: void ndv_dehomogenize(NDV p,struct order_spec *spec);
1.40 noro 358: void ndv_removecont(int mod,NDV p);
359: void ndv_print(NDV p);
360: void ndv_print_q(NDV p);
361: void ndv_free(NDV p);
1.77 noro 362: void ndv_save(NDV p,int index);
363: NDV ndv_load(int index);
1.40 noro 364:
365: /* converters */
1.61 noro 366: ND ptond(VL vl,VL dvl,P p);
367: NDV ptondv(VL vl,VL dvl,P p);
368: P ndvtop(int mod,VL vl,VL dvl,NDV p);
1.16 noro 369: NDV ndtondv(int mod,ND p);
1.23 noro 370: ND ndvtond(int mod,NDV p);
1.63 noro 371: int nm_ind_pair_to_vect(int m,UINT *s0,int n,NM_ind_pair pair,UINT *r);
1.67 noro 372: IndArray nm_ind_pair_to_vect_compress(int m,UINT *s0,int n,NM_ind_pair pair);
1.63 noro 373: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r);
1.1 noro 374:
1.76 noro 375: /* elimination */
376: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat);
377: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat);
378:
1.1 noro 379: void nd_free_private_storage()
380: {
381: _nm_free_list = 0;
1.5 noro 382: _ndp_free_list = 0;
1.71 noro 383: #if 0
1.1 noro 384: GC_gcollect();
1.71 noro 385: #endif
1.1 noro 386: }
387:
388: void _NM_alloc()
389: {
390: NM p;
391: int i;
392:
1.11 noro 393: for ( i = 0; i < 1024; i++ ) {
1.61 noro 394: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 395: p->next = _nm_free_list; _nm_free_list = p;
396: }
397: }
398:
399: void _ND_alloc()
400: {
401: ND p;
402: int i;
403:
404: for ( i = 0; i < 1024; i++ ) {
405: p = (ND)GC_malloc(sizeof(struct oND));
406: p->body = (NM)_nd_free_list; _nd_free_list = p;
407: }
408: }
409:
410: void _NDP_alloc()
411: {
412: ND_pairs p;
413: int i;
414:
1.11 noro 415: for ( i = 0; i < 1024; i++ ) {
1.1 noro 416: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 417: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 418: p->next = _ndp_free_list; _ndp_free_list = p;
419: }
420: }
421:
1.30 noro 422: INLINE int nd_length(ND p)
1.1 noro 423: {
424: NM m;
425: int i;
426:
427: if ( !p )
428: return 0;
429: else {
430: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
431: return i;
432: }
433: }
434:
1.61 noro 435: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 436: {
1.61 noro 437: UINT u1,u2;
1.1 noro 438: int i,j;
439:
1.34 noro 440: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 441: #if USE_UNROLL
1.1 noro 442: switch ( nd_bpe ) {
1.62 noro 443: case 3:
444: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
445: u1 = d1[i]; u2 = d2[i];
446: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
447: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
448: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
449: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
450: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
451: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
452: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
453: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
454: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
455: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
456: }
457: return 1;
458: break;
1.1 noro 459: case 4:
1.41 noro 460: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 461: u1 = d1[i]; u2 = d2[i];
462: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 463: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
464: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
465: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
466: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
467: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
468: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
469: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 470: }
471: return 1;
472: break;
473: case 6:
1.41 noro 474: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 475: u1 = d1[i]; u2 = d2[i];
476: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 477: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
478: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
479: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
480: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 481: }
482: return 1;
483: break;
484: case 8:
1.41 noro 485: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 486: u1 = d1[i]; u2 = d2[i];
487: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 488: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
489: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
490: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 491: }
492: return 1;
493: break;
494: case 16:
1.41 noro 495: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 496: u1 = d1[i]; u2 = d2[i];
497: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 498: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 499: }
500: return 1;
501: break;
502: case 32:
1.41 noro 503: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 504: if ( d1[i] < d2[i] ) return 0;
505: return 1;
506: break;
507: default:
1.41 noro 508: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 509: u1 = d1[i]; u2 = d2[i];
510: for ( j = 0; j < nd_epw; j++ )
511: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
512: }
513: return 1;
514: }
1.65 noro 515: #else
516: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
517: u1 = d1[i]; u2 = d2[i];
518: for ( j = 0; j < nd_epw; j++ )
519: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
520: }
521: return 1;
522: #endif
1.1 noro 523: }
524:
1.61 noro 525: /*
526: * If the current order is a block order,
527: * then the last block is length 1 and contains
528: * the homo variable. Otherwise, the original
529: * order is either 0 or 2.
530: */
531:
532: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 533: {
1.61 noro 534: int w,i,e,n,omask0;
535:
536: omask0 = (1<<obpe)-1;
537: n = nd_nvar-1;
538: ndl_zero(r);
539: for ( i = 0; i < n; i++ ) {
540: e = GET_EXP_OLD(d,i);
541: PUT_EXP(r,i,e);
542: }
543: w = TD(d);
544: PUT_EXP(r,nd_nvar-1,weight-w);
545: TD(r) = weight;
546: if ( nd_blockmask ) ndl_weight_mask(r);
547: }
548:
549: void ndl_dehomogenize(UINT *d)
550: {
551: UINT mask;
552: UINT h;
1.31 noro 553: int i,bits;
1.23 noro 554:
1.44 noro 555: if ( nd_blockmask ) {
556: h = GET_EXP(d,nd_nvar-1);
1.45 noro 557: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 558: TD(d) -= h;
559: d[nd_exporigin-1] -= h;
560: } else {
561: if ( nd_isrlex ) {
562: if ( nd_bpe == 32 ) {
563: h = d[nd_exporigin];
564: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
565: d[i-1] = d[i];
566: d[i-1] = 0;
567: TD(d) -= h;
568: } else {
569: bits = nd_epw*nd_bpe;
570: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
571: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
572: for ( i = nd_exporigin; i < nd_wpd; i++ )
573: d[i] = ((d[i]<<nd_bpe)&mask)
574: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
575: TD(d) -= h;
576: }
1.45 noro 577: } else {
578: h = GET_EXP(d,nd_nvar-1);
579: XOR_EXP(d,nd_nvar-1,h);
580: TD(d) -= h;
581: }
1.44 noro 582: }
1.23 noro 583: }
584:
1.61 noro 585: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 586: {
1.61 noro 587: UINT t1,t2,u,u1,u2;
1.43 noro 588: int i,j,l;
1.1 noro 589:
1.65 noro 590: #if USE_UNROLL
1.1 noro 591: switch ( nd_bpe ) {
1.62 noro 592: case 3:
593: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
594: u1 = d1[i]; u2 = d2[i];
595: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
596: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
597: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
598: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
599: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
600: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
601: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
602: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
603: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
604: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
605: d[i] = u;
606: }
607: break;
1.1 noro 608: case 4:
1.41 noro 609: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 610: u1 = d1[i]; u2 = d2[i];
611: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 612: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
613: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
614: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
615: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
616: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
617: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
618: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 619: d[i] = u;
620: }
621: break;
622: case 6:
1.41 noro 623: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 624: u1 = d1[i]; u2 = d2[i];
625: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 626: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
627: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
628: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
629: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 630: d[i] = u;
631: }
632: break;
633: case 8:
1.41 noro 634: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 635: u1 = d1[i]; u2 = d2[i];
636: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 637: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
638: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
639: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 640: d[i] = u;
641: }
642: break;
643: case 16:
1.41 noro 644: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 645: u1 = d1[i]; u2 = d2[i];
646: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 647: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 648: d[i] = u;
649: }
650: break;
651: case 32:
1.41 noro 652: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 653: u1 = d1[i]; u2 = d2[i];
654: d[i] = u1>u2?u1:u2;
655: }
656: break;
657: default:
1.41 noro 658: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 659: u1 = d1[i]; u2 = d2[i];
660: for ( j = 0, u = 0; j < nd_epw; j++ ) {
661: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
662: }
663: d[i] = u;
664: }
665: break;
666: }
1.65 noro 667: #else
668: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
669: u1 = d1[i]; u2 = d2[i];
670: for ( j = 0, u = 0; j < nd_epw; j++ ) {
671: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
672: }
673: d[i] = u;
674: }
675: #endif
1.39 noro 676: TD(d) = ndl_weight(d);
1.61 noro 677: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 678: }
679:
1.61 noro 680: int ndl_weight(UINT *d)
1.1 noro 681: {
1.61 noro 682: UINT t,u;
1.1 noro 683: int i,j;
684:
1.60 noro 685: if ( current_dl_weight_vector )
686: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
687: u = GET_EXP(d,i);
688: t += MUL_WEIGHT(u,i);
689: }
690: else
691: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
692: u = d[i];
693: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
694: t += (u&nd_mask0);
695: }
1.1 noro 696: return t;
697: }
698:
1.61 noro 699: void ndl_weight_mask(UINT *d)
1.43 noro 700: {
1.61 noro 701: UINT t,u;
702: UINT *mask;
703: int i,j,k,l;
1.43 noro 704:
1.61 noro 705: l = nd_blockmask->n;
706: for ( k = 0; k < l; k++ ) {
707: mask = nd_blockmask->mask[k];
708: if ( current_dl_weight_vector )
709: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
710: u = GET_EXP_MASK(d,i,mask);
711: t += MUL_WEIGHT(u,i);
712: }
713: else
714: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
715: u = d[i]&mask[i];
716: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
717: t += (u&nd_mask0);
718: }
719: d[k+1] = t;
720: }
1.43 noro 721: }
722:
1.61 noro 723: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 724: {
725: int i;
726:
1.41 noro 727: d1 += nd_exporigin;
728: d2 += nd_exporigin;
729: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 730: if ( *d1 > *d2 )
1.32 noro 731: return nd_isrlex ? -1 : 1;
1.1 noro 732: else if ( *d1 < *d2 )
1.32 noro 733: return nd_isrlex ? 1 : -1;
1.1 noro 734: return 0;
735: }
736:
1.61 noro 737: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 738: {
739: int i,l,j,ord_o,ord_l;
740: struct order_pair *op;
1.61 noro 741: UINT t1,t2,m;
742: UINT *mask;
1.43 noro 743:
744: l = nd_blockmask->n;
745: op = nd_blockmask->order_pair;
746: for ( j = 0; j < l; j++ ) {
747: mask = nd_blockmask->mask[j];
748: ord_o = op[j].order;
749: if ( ord_o < 2 )
1.44 noro 750: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
751: else if ( t1 < t2 ) return -1;
1.43 noro 752: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 753: m = mask[i];
754: t1 = d1[i]&m;
755: t2 = d2[i]&m;
1.43 noro 756: if ( t1 > t2 )
757: return !ord_o ? -1 : 1;
758: else if ( t1 < t2 )
759: return !ord_o ? 1 : -1;
760: }
761: }
762: return 0;
763: }
764:
1.58 noro 765: /* TDH -> WW -> TD-> RL */
766:
1.61 noro 767: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 768: {
769: int i,m,e1,e2;
770:
771: if ( TD(d1) > TD(d2) ) return 1;
772: else if ( TD(d1) < TD(d2) ) return -1;
773: m = nd_nvar>>1;
774: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
775: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
776: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
777: }
778: if ( e1 > e2 ) return 1;
779: else if ( e1 < e2 ) return -1;
780: return ndl_lex_compare(d1,d2);
781: }
782:
1.61 noro 783: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 784: {
785: int i;
786:
1.81 noro 787: switch ( nd_wpd ) {
788: case 2:
789: if ( TD(d2) != TD(d1) ) return 0;
790: if ( d2[1] != d1[1] ) return 0;
791: return 1;
792: break;
793: case 3:
794: if ( TD(d2) != TD(d1) ) return 0;
795: if ( d2[1] != d1[1] ) return 0;
796: if ( d2[2] != d1[2] ) return 0;
797: return 1;
798: break;
799: default:
800: for ( i = 0; i < nd_wpd; i++ )
801: if ( *d1++ != *d2++ ) return 0;
802: return 1;
803: break;
804: }
1.1 noro 805: }
806:
1.61 noro 807: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 808: {
809: int i;
810:
811: switch ( nd_wpd ) {
1.41 noro 812: case 2:
1.34 noro 813: TD(d2) = TD(d1);
814: d2[1] = d1[1];
1.6 noro 815: break;
1.41 noro 816: case 3:
1.34 noro 817: TD(d2) = TD(d1);
1.6 noro 818: d2[1] = d1[1];
1.34 noro 819: d2[2] = d1[2];
1.6 noro 820: break;
821: default:
1.41 noro 822: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 823: d2[i] = d1[i];
824: break;
825: }
826: }
827:
1.61 noro 828: INLINE void ndl_zero(UINT *d)
829: {
830: int i;
831: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
832: }
833:
834: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 835: {
836: int i;
837:
1.43 noro 838: #if 1
1.6 noro 839: switch ( nd_wpd ) {
1.41 noro 840: case 2:
841: TD(d) = TD(d1)+TD(d2);
1.34 noro 842: d[1] = d1[1]+d2[1];
1.6 noro 843: break;
1.41 noro 844: case 3:
845: TD(d) = TD(d1)+TD(d2);
1.6 noro 846: d[1] = d1[1]+d2[1];
1.34 noro 847: d[2] = d1[2]+d2[2];
1.6 noro 848: break;
849: default:
1.43 noro 850: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 851: break;
852: }
1.43 noro 853: #else
854: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
855: #endif
1.6 noro 856: }
857:
1.55 noro 858: /* d1 += d2 */
1.61 noro 859: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 860: {
861: int i;
862:
863: #if 1
864: switch ( nd_wpd ) {
865: case 2:
866: TD(d1) += TD(d2);
867: d1[1] += d2[1];
868: break;
869: case 3:
870: TD(d1) += TD(d2);
871: d1[1] += d2[1];
872: d1[2] += d2[2];
873: break;
874: default:
875: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
876: break;
877: }
878: #else
879: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
880: #endif
881: }
882:
1.61 noro 883: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 884: {
885: int i;
886:
1.43 noro 887: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 888: }
889:
1.61 noro 890: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 891: {
1.61 noro 892: UINT t1,t2,u,u1,u2;
1.1 noro 893: int i,j;
894:
1.65 noro 895: #if USE_UNROLL
1.1 noro 896: switch ( nd_bpe ) {
1.62 noro 897: case 3:
898: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
899: u1 = d1[i]; u2 = d2[i];
900: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
901: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
902: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
903: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
904: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
905: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
906: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
907: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
908: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
909: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
910: }
911: return 1;
912: break;
1.1 noro 913: case 4:
1.41 noro 914: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 915: u1 = d1[i]; u2 = d2[i];
916: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 917: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
918: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
919: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
920: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
921: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
922: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
923: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 924: }
925: return 1;
926: break;
927: case 6:
1.41 noro 928: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 929: u1 = d1[i]; u2 = d2[i];
930: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 931: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
932: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
933: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
934: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 935: }
936: return 1;
937: break;
938: case 8:
1.41 noro 939: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 940: u1 = d1[i]; u2 = d2[i];
941: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 942: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
943: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
944: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 945: }
946: return 1;
947: break;
948: case 16:
1.41 noro 949: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 950: u1 = d1[i]; u2 = d2[i];
951: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 952: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 953: }
954: return 1;
955: break;
956: case 32:
1.41 noro 957: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 958: if ( d1[i] && d2[i] ) return 0;
959: return 1;
960: break;
961: default:
1.41 noro 962: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 963: u1 = d1[i]; u2 = d2[i];
964: for ( j = 0; j < nd_epw; j++ ) {
965: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
966: u1 >>= nd_bpe; u2 >>= nd_bpe;
967: }
968: }
969: return 1;
970: break;
971: }
1.65 noro 972: #else
973: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
974: u1 = d1[i]; u2 = d2[i];
975: for ( j = 0; j < nd_epw; j++ ) {
976: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
977: u1 >>= nd_bpe; u2 >>= nd_bpe;
978: }
979: }
980: return 1;
981: #endif
1.1 noro 982: }
983:
1.61 noro 984: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 985: {
1.61 noro 986: UINT u2;
987: UINT *d1;
1.5 noro 988: int i,j,ind,k;
1.1 noro 989:
1.5 noro 990: d1 = nd_bound[index];
991: ind = 0;
1.65 noro 992: #if USE_UNROLL
1.5 noro 993: switch ( nd_bpe ) {
1.62 noro 994: case 3:
995: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
996: u2 = d2[i];
997: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
998: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
999: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
1000: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
1001: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
1002: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
1003: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
1004: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
1005: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
1006: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
1007: }
1008: return 0;
1009: break;
1.5 noro 1010: case 4:
1.41 noro 1011: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1012: u2 = d2[i];
1013: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
1014: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
1015: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
1016: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
1017: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
1018: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
1019: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1020: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1021: }
1022: return 0;
1023: break;
1024: case 6:
1.41 noro 1025: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1026: u2 = d2[i];
1027: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1028: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1029: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1030: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1031: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1032: }
1033: return 0;
1034: break;
1035: case 8:
1.41 noro 1036: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1037: u2 = d2[i];
1038: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1039: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1040: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1041: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1042: }
1043: return 0;
1044: break;
1045: case 16:
1.41 noro 1046: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1047: u2 = d2[i];
1048: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1049: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1050: }
1051: return 0;
1052: break;
1053: case 32:
1.41 noro 1054: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1055: if ( d1[i]+d2[i]<d1[i] ) return 1;
1056: return 0;
1057: break;
1058: default:
1.41 noro 1059: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1060: u2 = d2[i];
1061: k = (nd_epw-1)*nd_bpe;
1062: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1063: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1064: }
1065: return 0;
1066: break;
1067: }
1.65 noro 1068: #else
1069: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1070: u2 = d2[i];
1071: k = (nd_epw-1)*nd_bpe;
1072: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1073: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1074: }
1075: return 0;
1076: #endif
1.1 noro 1077: }
1078:
1.61 noro 1079: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1080: {
1081: int i;
1082: int r;
1083:
1.34 noro 1084: r = 0;
1.41 noro 1085: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1086: r = ((r<<16)+d[i])%REDTAB_LEN;
1087: return r;
1088: }
1089:
1.63 noro 1090: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1091: {
1.13 noro 1092: RHist r;
1.6 noro 1093: int d,k,i;
1.1 noro 1094:
1.63 noro 1095: d = ndl_hash_value(dg);
1.13 noro 1096: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1097: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1098: if ( k > 0 ) nd_notfirst++;
1099: nd_found++;
1.13 noro 1100: return r->index;
1.1 noro 1101: }
1102: }
1.13 noro 1103: if ( Reverse )
1104: for ( i = nd_psn-1; i >= 0; i-- ) {
1105: r = nd_psh[i];
1.34 noro 1106: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1107: nd_create++;
1.34 noro 1108: nd_append_red(dg,i);
1.13 noro 1109: return i;
1110: }
1111: }
1112: else
1113: for ( i = 0; i < nd_psn; i++ ) {
1114: r = nd_psh[i];
1.34 noro 1115: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1116: nd_create++;
1.34 noro 1117: nd_append_red(dg,i);
1.13 noro 1118: return i;
1119: }
1.1 noro 1120: }
1.6 noro 1121: return -1;
1.1 noro 1122: }
1123:
1.63 noro 1124: ND nd_merge(ND p1,ND p2)
1125: {
1126: int n,c;
1127: int t,can,td1,td2;
1128: ND r;
1129: NM m1,m2,mr0,mr,s;
1130:
1131: if ( !p1 ) return p2;
1132: else if ( !p2 ) return p1;
1133: else {
1134: can = 0;
1135: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1136: c = DL_COMPARE(DL(m1),DL(m2));
1137: switch ( c ) {
1138: case 0:
1139: s = m1; m1 = NEXT(m1);
1140: can++; NEXTNM2(mr0,mr,s);
1141: s = m2; m2 = NEXT(m2); FREENM(s);
1142: break;
1143: case 1:
1144: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1145: break;
1146: case -1:
1147: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1148: break;
1149: }
1150: }
1151: if ( !mr0 )
1152: if ( m1 ) mr0 = m1;
1153: else if ( m2 ) mr0 = m2;
1154: else return 0;
1155: else if ( m1 ) NEXT(mr) = m1;
1156: else if ( m2 ) NEXT(mr) = m2;
1157: else NEXT(mr) = 0;
1158: BDY(p1) = mr0;
1159: SG(p1) = MAX(SG(p1),SG(p2));
1160: LEN(p1) = LEN(p1)+LEN(p2)-can;
1161: FREEND(p2);
1162: return p1;
1163: }
1164: }
1165:
1.31 noro 1166: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1167: {
1168: int n,c;
1.34 noro 1169: int t,can,td1,td2;
1.1 noro 1170: ND r;
1171: NM m1,m2,mr0,mr,s;
1172:
1.34 noro 1173: if ( !p1 ) return p2;
1174: else if ( !p2 ) return p1;
1.71 noro 1175: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 1176: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1177: else {
1.30 noro 1178: can = 0;
1.1 noro 1179: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1180: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1181: switch ( c ) {
1182: case 0:
1.19 noro 1183: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1184: if ( t < 0 ) t += mod;
1.1 noro 1185: s = m1; m1 = NEXT(m1);
1186: if ( t ) {
1.34 noro 1187: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1188: } else {
1.34 noro 1189: can += 2; FREENM(s);
1.1 noro 1190: }
1191: s = m2; m2 = NEXT(m2); FREENM(s);
1192: break;
1193: case 1:
1194: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1195: break;
1196: case -1:
1197: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1198: break;
1199: }
1200: }
1201: if ( !mr0 )
1.34 noro 1202: if ( m1 ) mr0 = m1;
1203: else if ( m2 ) mr0 = m2;
1204: else return 0;
1205: else if ( m1 ) NEXT(mr) = m1;
1206: else if ( m2 ) NEXT(mr) = m2;
1207: else NEXT(mr) = 0;
1.1 noro 1208: BDY(p1) = mr0;
1.14 noro 1209: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1210: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1211: FREEND(p2);
1212: return p1;
1213: }
1214: }
1215:
1.31 noro 1216: ND nd_add_q(ND p1,ND p2)
1.17 noro 1217: {
1.30 noro 1218: int n,c,can;
1.17 noro 1219: ND r;
1220: NM m1,m2,mr0,mr,s;
1221: Q t;
1222:
1.34 noro 1223: if ( !p1 ) return p2;
1224: else if ( !p2 ) return p1;
1.31 noro 1225: else {
1.30 noro 1226: can = 0;
1.17 noro 1227: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1228: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1229: switch ( c ) {
1230: case 0:
1231: addq(CQ(m1),CQ(m2),&t);
1232: s = m1; m1 = NEXT(m1);
1233: if ( t ) {
1.34 noro 1234: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1235: } else {
1.34 noro 1236: can += 2; FREENM(s);
1.17 noro 1237: }
1238: s = m2; m2 = NEXT(m2); FREENM(s);
1239: break;
1240: case 1:
1241: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1242: break;
1243: case -1:
1244: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1245: break;
1246: }
1247: }
1248: if ( !mr0 )
1.34 noro 1249: if ( m1 ) mr0 = m1;
1250: else if ( m2 ) mr0 = m2;
1251: else return 0;
1252: else if ( m1 ) NEXT(mr) = m1;
1253: else if ( m2 ) NEXT(mr) = m2;
1254: else NEXT(mr) = 0;
1.17 noro 1255: BDY(p1) = mr0;
1256: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1257: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1258: FREEND(p2);
1259: return p1;
1260: }
1261: }
1262:
1.71 noro 1263: ND nd_add_sf(ND p1,ND p2)
1264: {
1265: int n,c,can;
1266: ND r;
1267: NM m1,m2,mr0,mr,s;
1268: int t;
1269:
1270: if ( !p1 ) return p2;
1271: else if ( !p2 ) return p1;
1272: else {
1273: can = 0;
1274: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1275: c = DL_COMPARE(DL(m1),DL(m2));
1276: switch ( c ) {
1277: case 0:
1278: t = _addsf(CM(m1),CM(m2));
1279: s = m1; m1 = NEXT(m1);
1280: if ( t ) {
1281: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1282: } else {
1283: can += 2; FREENM(s);
1284: }
1285: s = m2; m2 = NEXT(m2); FREENM(s);
1286: break;
1287: case 1:
1288: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1289: break;
1290: case -1:
1291: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1292: break;
1293: }
1294: }
1295: if ( !mr0 )
1296: if ( m1 ) mr0 = m1;
1297: else if ( m2 ) mr0 = m2;
1298: else return 0;
1299: else if ( m1 ) NEXT(mr) = m1;
1300: else if ( m2 ) NEXT(mr) = m2;
1301: else NEXT(mr) = 0;
1302: BDY(p1) = mr0;
1303: SG(p1) = MAX(SG(p1),SG(p2));
1304: LEN(p1) = LEN(p1)+LEN(p2)-can;
1305: FREEND(p2);
1306: return p1;
1307: }
1308: }
1309:
1.1 noro 1310: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1311: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1312: {
1.11 noro 1313: ND d;
1.1 noro 1314: NM m,mrd,tail;
1.7 noro 1315: NM mul;
1.10 noro 1316: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1317: int c,c1,c2,dummy;
1.17 noro 1318: RHist h;
1.11 noro 1319: NDV p,red;
1.69 noro 1320: Q cg,cred,gcd,tq,qq;
1.21 noro 1321: double hmag;
1.1 noro 1322:
1.69 noro 1323: if ( dn ) {
1324: if ( mod )
1325: dn->m = 1;
1326: else
1327: dn->z = ONE;
1328: }
1.1 noro 1329: if ( !g ) {
1330: *rp = 0;
1331: return 1;
1332: }
1.34 noro 1333: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1334:
1.14 noro 1335: sugar0 = sugar = SG(g);
1.1 noro 1336: n = NV(g);
1.61 noro 1337: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1338: for ( d = 0; g; ) {
1.63 noro 1339: index = ndl_find_reducer(HDL(g));
1.6 noro 1340: if ( index >= 0 ) {
1.17 noro 1341: h = nd_psh[index];
1342: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1343: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1344: nd_free(g); nd_free(d);
1345: return 0;
1346: }
1.77 noro 1347: if ( nd_demand )
1348: p = ndv_load(index);
1349: else
1350: p = ps[index];
1.71 noro 1351: if ( mod == -1 )
1352: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1353: else if ( mod ) {
1.19 noro 1354: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1355: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1356: } else {
1.17 noro 1357: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1358: chsgnq(cg,&CQ(mul));
1.20 noro 1359: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1360: if ( dn ) {
1361: mulq(dn->z,cred,&tq); dn->z = tq;
1362: }
1.16 noro 1363: }
1.55 noro 1364: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1365: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1366: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1367: tq = HCQ(g);
1.21 noro 1368: nd_removecont2(d,g);
1.69 noro 1369: if ( dn ) {
1370: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1371: }
1.21 noro 1372: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1373: }
1.1 noro 1374: } else if ( !full ) {
1375: *rp = g;
1376: return 1;
1377: } else {
1378: m = BDY(g);
1379: if ( NEXT(m) ) {
1.34 noro 1380: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1381: } else {
1382: FREEND(g); g = 0;
1383: }
1384: if ( d ) {
1.34 noro 1385: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1386: } else {
1.34 noro 1387: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1388: }
1389: }
1390: }
1.34 noro 1391: if ( d ) SG(d) = sugar;
1.1 noro 1392: *rp = d;
1393: return 1;
1394: }
1.28 noro 1395:
1.53 noro 1396: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1397: {
1398: int hindex,index;
1399: NDV p;
1400: ND u,d,red;
1401: NODE l;
1.31 noro 1402: NM mul,m,mrd,tail;
1.25 noro 1403: int sugar,psugar,n,h_reducible;
1404: PGeoBucket bucket;
1405: int c,c1,c2;
1.26 noro 1406: Q cg,cred,gcd,zzz;
1.25 noro 1407: RHist h;
1.28 noro 1408: double hmag,gmag;
1.77 noro 1409: int count = 0;
1410: int hcount = 0;
1.25 noro 1411:
1412: if ( !g ) {
1413: *rp = 0;
1414: return 1;
1415: }
1416: sugar = SG(g);
1417: n = NV(g);
1.34 noro 1418: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1419: bucket = create_pbucket();
1.31 noro 1420: add_pbucket(mod,bucket,g);
1.25 noro 1421: d = 0;
1.61 noro 1422: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1423: while ( 1 ) {
1.26 noro 1424: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1425: if ( hindex < 0 ) {
1.77 noro 1426: if ( DP_Print > 3 ) printf("(%d %d)",count,hcount);
1.34 noro 1427: if ( d ) SG(d) = sugar;
1.25 noro 1428: *rp = d;
1429: return 1;
1430: }
1431: g = bucket->body[hindex];
1.63 noro 1432: index = ndl_find_reducer(HDL(g));
1.25 noro 1433: if ( index >= 0 ) {
1.77 noro 1434: count++;
1435: if ( !d ) hcount++;
1.25 noro 1436: h = nd_psh[index];
1437: ndl_sub(HDL(g),DL(h),DL(mul));
1438: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1439: nd_free(d);
1.25 noro 1440: free_pbucket(bucket);
1441: *rp = 0;
1442: return 0;
1443: }
1.53 noro 1444: p = ps[index];
1.71 noro 1445: if ( mod == -1 )
1446: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1447: else if ( mod ) {
1.25 noro 1448: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1449: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1450: } else {
1451: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1452: chsgnq(cg,&CQ(mul));
1.26 noro 1453: nd_mul_c_q(d,cred);
1454: mulq_pbucket(bucket,cred);
1455: g = bucket->body[hindex];
1.28 noro 1456: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1457: }
1.55 noro 1458: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1459: bucket->body[hindex] = nd_remove_head(g);
1460: red = nd_remove_head(red);
1.31 noro 1461: add_pbucket(mod,bucket,red);
1.34 noro 1462: psugar = SG(p)+TD(DL(mul));
1463: sugar = MAX(sugar,psugar);
1.28 noro 1464: if ( !mod && hmag && (gmag > hmag) ) {
1465: g = normalize_pbucket(mod,bucket);
1466: if ( !g ) {
1.34 noro 1467: if ( d ) SG(d) = sugar;
1.28 noro 1468: *rp = d;
1469: return 1;
1470: }
1471: nd_removecont2(d,g);
1472: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1473: add_pbucket(mod,bucket,g);
1.28 noro 1474: }
1.25 noro 1475: } else if ( !full ) {
1476: g = normalize_pbucket(mod,bucket);
1.34 noro 1477: if ( g ) SG(g) = sugar;
1.25 noro 1478: *rp = g;
1479: return 1;
1480: } else {
1481: m = BDY(g);
1482: if ( NEXT(m) ) {
1.34 noro 1483: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1484: } else {
1485: FREEND(g); g = 0;
1486: }
1487: bucket->body[hindex] = g;
1488: NEXT(m) = 0;
1489: if ( d ) {
1.34 noro 1490: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1491: } else {
1.34 noro 1492: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1493: }
1494: }
1495: }
1496: }
1.27 noro 1497:
1.61 noro 1498: /* input : list of NDV, cand : list of NDV */
1.28 noro 1499:
1.61 noro 1500: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1501: {
1502: int n,i,stat;
1503: ND nf,d;
1.61 noro 1504: NDV r;
1.45 noro 1505: NODE t,s;
1.69 noro 1506: union oNDC dn;
1.45 noro 1507:
1.61 noro 1508: ndv_setup(0,0,cand);
1.31 noro 1509: n = length(cand);
1.28 noro 1510:
1511: /* membercheck : list is a subset of Id(cand) ? */
1512: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1513: again:
1.61 noro 1514: if ( nd_bpe > obpe )
1515: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1516: else
1517: r = (NDV)BDY(t);
1518: d = ndvtond(0,r);
1.69 noro 1519: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1520: if ( !stat ) {
1521: nd_reconstruct(0,0,0);
1522: goto again;
1523: } else if ( nf ) return 0;
1.71 noro 1524: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1525: }
1.71 noro 1526: if ( DP_Print ) { printf("\n"); }
1.28 noro 1527: /* gbcheck : cand is a GB of Id(cand) ? */
1.77 noro 1528: if ( !nd_gb(0,0,1) ) return 0;
1.28 noro 1529: /* XXX */
1.23 noro 1530: return 1;
1531: }
1.1 noro 1532:
1533: ND nd_remove_head(ND p)
1534: {
1535: NM m;
1536:
1537: m = BDY(p);
1538: if ( !NEXT(m) ) {
1.34 noro 1539: FREEND(p); p = 0;
1.31 noro 1540: } else {
1.34 noro 1541: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1542: }
1.1 noro 1543: FREENM(m);
1544: return p;
1545: }
1546:
1.69 noro 1547: ND nd_separate_head(ND p,ND *head)
1548: {
1549: NM m,m0;
1550: ND r;
1551:
1552: m = BDY(p);
1553: if ( !NEXT(m) ) {
1554: *head = p; p = 0;
1555: } else {
1556: m0 = m;
1557: BDY(p) = NEXT(m); LEN(p)--;
1558: NEXT(m0) = 0;
1559: MKND(NV(p),m0,1,r);
1560: *head = r;
1561: }
1562: return p;
1563: }
1564:
1.1 noro 1565: PGeoBucket create_pbucket()
1566: {
1567: PGeoBucket g;
1568:
1569: g = CALLOC(1,sizeof(struct oPGeoBucket));
1570: g->m = -1;
1571: return g;
1572: }
1573:
1.25 noro 1574: void free_pbucket(PGeoBucket b) {
1575: int i;
1576:
1.26 noro 1577: for ( i = 0; i <= b->m; i++ )
1.25 noro 1578: if ( b->body[i] ) {
1579: nd_free(b->body[i]);
1580: b->body[i] = 0;
1581: }
1582: GC_free(b);
1583: }
1584:
1.63 noro 1585: void add_pbucket_symbolic(PGeoBucket g,ND d)
1586: {
1587: int l,i,k,m;
1588:
1589: if ( !d )
1590: return;
1591: l = LEN(d);
1592: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1593: /* 2^(k-1) < l <= 2^k (=m) */
1594: d = nd_merge(g->body[k],d);
1595: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1596: g->body[k] = 0;
1597: d = nd_merge(g->body[k+1],d);
1598: }
1599: g->body[k] = d;
1600: g->m = MAX(g->m,k);
1601: }
1602:
1.31 noro 1603: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1604: {
1.31 noro 1605: int l,i,k,m;
1.1 noro 1606:
1.31 noro 1607: if ( !d )
1608: return;
1609: l = LEN(d);
1.29 noro 1610: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1611: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1612: d = nd_add(mod,g->body[k],d);
1613: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1614: g->body[k] = 0;
1.31 noro 1615: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1616: }
1617: g->body[k] = d;
1618: g->m = MAX(g->m,k);
1619: }
1620:
1.26 noro 1621: void mulq_pbucket(PGeoBucket g,Q c)
1622: {
1623: int k;
1624:
1625: for ( k = 0; k <= g->m; k++ )
1626: nd_mul_c_q(g->body[k],c);
1627: }
1628:
1.63 noro 1629: NM remove_head_pbucket_symbolic(PGeoBucket g)
1630: {
1631: int j,i,k,c;
1632: NM head;
1633:
1634: k = g->m;
1635: j = -1;
1636: for ( i = 0; i <= k; i++ ) {
1637: if ( !g->body[i] ) continue;
1638: if ( j < 0 ) j = i;
1639: else {
1640: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1641: if ( c > 0 )
1642: j = i;
1643: else if ( c == 0 )
1644: g->body[i] = nd_remove_head(g->body[i]);
1645: }
1646: }
1647: if ( j < 0 ) return 0;
1648: else {
1649: head = BDY(g->body[j]);
1650: if ( !NEXT(head) ) {
1651: FREEND(g->body[j]);
1652: g->body[j] = 0;
1653: } else {
1654: BDY(g->body[j]) = NEXT(head);
1655: LEN(g->body[j])--;
1656: }
1657: return head;
1658: }
1659: }
1660:
1.19 noro 1661: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1662: {
1663: int j,i,c,k,nv,sum;
1.61 noro 1664: UINT *di,*dj;
1.1 noro 1665: ND gi,gj;
1666:
1667: k = g->m;
1668: while ( 1 ) {
1669: j = -1;
1670: for ( i = 0; i <= k; i++ ) {
1671: if ( !(gi = g->body[i]) )
1672: continue;
1673: if ( j < 0 ) {
1674: j = i;
1675: gj = g->body[j];
1676: dj = HDL(gj);
1.14 noro 1677: sum = HCM(gj);
1.1 noro 1678: } else {
1.34 noro 1679: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1680: if ( c > 0 ) {
1.34 noro 1681: if ( sum ) HCM(gj) = sum;
1682: else g->body[j] = nd_remove_head(gj);
1.1 noro 1683: j = i;
1684: gj = g->body[j];
1685: dj = HDL(gj);
1.14 noro 1686: sum = HCM(gj);
1.1 noro 1687: } else if ( c == 0 ) {
1.71 noro 1688: if ( mod == -1 )
1689: sum = _addsf(sum,HCM(gi));
1690: else {
1691: sum = sum+HCM(gi)-mod;
1692: if ( sum < 0 ) sum += mod;
1693: }
1.1 noro 1694: g->body[i] = nd_remove_head(gi);
1695: }
1696: }
1697: }
1.34 noro 1698: if ( j < 0 ) return -1;
1.1 noro 1699: else if ( sum ) {
1.14 noro 1700: HCM(gj) = sum;
1.26 noro 1701: return j;
1.31 noro 1702: } else
1.26 noro 1703: g->body[j] = nd_remove_head(gj);
1704: }
1705: }
1706:
1707: int head_pbucket_q(PGeoBucket g)
1708: {
1709: int j,i,c,k,nv;
1710: Q sum,t;
1711: ND gi,gj;
1712:
1713: k = g->m;
1714: while ( 1 ) {
1715: j = -1;
1716: for ( i = 0; i <= k; i++ ) {
1.34 noro 1717: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1718: if ( j < 0 ) {
1719: j = i;
1720: gj = g->body[j];
1721: sum = HCQ(gj);
1722: } else {
1723: nv = NV(gi);
1.34 noro 1724: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1725: if ( c > 0 ) {
1.34 noro 1726: if ( sum ) HCQ(gj) = sum;
1727: else g->body[j] = nd_remove_head(gj);
1.26 noro 1728: j = i;
1729: gj = g->body[j];
1730: sum = HCQ(gj);
1731: } else if ( c == 0 ) {
1732: addq(sum,HCQ(gi),&t);
1733: sum = t;
1734: g->body[i] = nd_remove_head(gi);
1735: }
1736: }
1737: }
1.34 noro 1738: if ( j < 0 ) return -1;
1.26 noro 1739: else if ( sum ) {
1740: HCQ(gj) = sum;
1.1 noro 1741: return j;
1.31 noro 1742: } else
1.1 noro 1743: g->body[j] = nd_remove_head(gj);
1744: }
1745: }
1746:
1.25 noro 1747: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1748: {
1.31 noro 1749: int i;
1.1 noro 1750: ND r,t;
1751:
1752: r = 0;
1.28 noro 1753: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1754: r = nd_add(mod,r,g->body[i]);
1.28 noro 1755: g->body[i] = 0;
1756: }
1757: g->m = -1;
1.1 noro 1758: return r;
1759: }
1760:
1.27 noro 1761: /* return value = 0 => input is not a GB */
1762:
1.77 noro 1763: NODE nd_gb(int m,int ishomo,int checkonly)
1.1 noro 1764: {
1765: int i,nh,sugar,stat;
1.23 noro 1766: NODE r,g,t;
1.1 noro 1767: ND_pairs d;
1768: ND_pairs l;
1.77 noro 1769: ND h,nf,s,head;
1.63 noro 1770: NDV nfv;
1.77 noro 1771: Q q;
1.69 noro 1772: union oNDC dn;
1.1 noro 1773:
1.23 noro 1774: g = 0; d = 0;
1775: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1776: d = update_pairs(d,g,i);
1777: g = update_base(g,i);
1778: }
1779: sugar = 0;
1780: while ( d ) {
1781: again:
1782: l = nd_minp(d,&d);
1.14 noro 1783: if ( SG(l) != sugar ) {
1.77 noro 1784: if ( ishomo ) {
1.82 ! noro 1785: for ( i = nd_psn-1; SG(nd_psh[i]) == sugar; i-- ) {
! 1786: if ( nd_demand )
! 1787: nfv = ndv_load(i);
! 1788: else
! 1789: nfv = nd_ps[i];
! 1790: s = ndvtond(m,nfv);
1.77 noro 1791: s = nd_separate_head(s,&head);
1792: nd_nf(m,s,nd_ps,1,&dn,&nf);
1793: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1794: nf = nd_add(m,head,nf);
1.82 ! noro 1795: ndv_free(nfv);
1.77 noro 1796: nd_removecont(m,nf);
1.82 ! noro 1797: nfv = ndtondv(m,nf);
! 1798: nd_free(nf);
! 1799: nd_bound[i] = ndv_compute_bound(nfv);
! 1800: if ( nd_demand ) {
! 1801: ndv_save(nfv,i);
! 1802: ndv_free(nfv);
! 1803: } else
! 1804: nd_ps[i] = nfv;
1.77 noro 1805: }
1806: }
1.14 noro 1807: sugar = SG(l);
1.71 noro 1808: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1809: }
1.53 noro 1810: stat = nd_sp(m,0,l,&h);
1.1 noro 1811: if ( !stat ) {
1812: NEXT(l) = d; d = l;
1.20 noro 1813: d = nd_reconstruct(m,0,d);
1.1 noro 1814: goto again;
1815: }
1.41 noro 1816: #if USE_GEOBUCKET
1.69 noro 1817: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1818: #else
1.69 noro 1819: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1820: #endif
1.1 noro 1821: if ( !stat ) {
1822: NEXT(l) = d; d = l;
1.20 noro 1823: d = nd_reconstruct(m,0,d);
1.1 noro 1824: goto again;
1825: } else if ( nf ) {
1.27 noro 1826: if ( checkonly ) return 0;
1.71 noro 1827: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1828: nd_removecont(m,nf);
1829: nfv = ndtondv(m,nf); nd_free(nf);
1.77 noro 1830: nh = ndv_newps(m,nfv,0);
1.1 noro 1831: d = update_pairs(d,g,nh);
1832: g = update_base(g,nh);
1833: FREENDP(l);
1834: } else {
1.71 noro 1835: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1836: FREENDP(l);
1837: }
1838: }
1.77 noro 1839: if ( nd_demand )
1840: for ( t = g; t; t = NEXT(t) )
1841: BDY(t) = (pointer)ndv_load((int)BDY(t));
1842: else
1843: for ( t = g; t; t = NEXT(t) )
1844: BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1845: return g;
1846: }
1847:
1.82 ! noro 1848: NODE nd_gb_trace(int m,int ishomo)
1.20 noro 1849: {
1850: int i,nh,sugar,stat;
1.23 noro 1851: NODE r,g,t;
1.20 noro 1852: ND_pairs d;
1853: ND_pairs l;
1.82 ! noro 1854: ND h,nf,nfq,s,head;
1.63 noro 1855: NDV nfv,nfqv;
1.82 ! noro 1856: Q q;
1.69 noro 1857: union oNDC dn;
1.20 noro 1858:
1.23 noro 1859: g = 0; d = 0;
1860: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1861: d = update_pairs(d,g,i);
1862: g = update_base(g,i);
1863: }
1864: sugar = 0;
1865: while ( d ) {
1866: again:
1867: l = nd_minp(d,&d);
1868: if ( SG(l) != sugar ) {
1.82 ! noro 1869: if ( ishomo ) {
! 1870: for ( i = nd_psn-1; SG(nd_psh[i]) == sugar; i-- ) {
! 1871: /* for nd_ps */
! 1872: s = ndvtond(m,nd_ps[i]);
! 1873: s = nd_separate_head(s,&head);
! 1874: nd_nf_pbucket(m,s,nd_ps,1,&nf);
! 1875: nf = nd_add(m,head,nf);
! 1876: ndv_free(nd_ps[i]);
! 1877: nd_ps[i] = ndtondv(m,nf);
! 1878: nd_free(nf);
! 1879:
! 1880: /* for nd_ps_trace */
! 1881: if ( nd_demand )
! 1882: nfv = ndv_load(i);
! 1883: else
! 1884: nfv = nd_ps_trace[i];
! 1885: s = ndvtond(0,nfv);
! 1886: s = nd_separate_head(s,&head);
! 1887: nd_nf(0,s,nd_ps_trace,1,&dn,&nf);
! 1888: mulq(HCQ(head),dn.z,&q); HCQ(head) = q;
! 1889: nf = nd_add(0,head,nf);
! 1890: ndv_free(nfv);
! 1891: nd_removecont(0,nf);
! 1892: nfv = ndtondv(0,nf);
! 1893: nd_free(nf);
! 1894: nd_bound[i] = ndv_compute_bound(nfv);
! 1895: if ( nd_demand ) {
! 1896: ndv_save(nfv,i);
! 1897: ndv_free(nfv);
! 1898: } else
! 1899: nd_ps_trace[i] = nfv;
! 1900: }
! 1901: }
1.20 noro 1902: sugar = SG(l);
1.71 noro 1903: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1904: }
1.53 noro 1905: stat = nd_sp(m,0,l,&h);
1.20 noro 1906: if ( !stat ) {
1907: NEXT(l) = d; d = l;
1908: d = nd_reconstruct(m,1,d);
1909: goto again;
1910: }
1.41 noro 1911: #if USE_GEOBUCKET
1.53 noro 1912: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1913: #else
1.69 noro 1914: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1915: #endif
1.20 noro 1916: if ( !stat ) {
1917: NEXT(l) = d; d = l;
1918: d = nd_reconstruct(m,1,d);
1919: goto again;
1920: } else if ( nf ) {
1.77 noro 1921: if ( nd_demand ) {
1922: nfqv = ndv_load(nd_psn);
1923: nfq = ndvtond(0,nfqv);
1924: } else
1925: nfq = 0;
1926: if ( !nfq ) {
1927: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1928: NEXT(l) = d; d = l;
1929: d = nd_reconstruct(m,1,d);
1930: goto again;
1931: }
1.70 noro 1932: }
1.20 noro 1933: if ( nfq ) {
1.70 noro 1934: /* m|HC(nfq) => failure */
1.63 noro 1935: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1936:
1.71 noro 1937: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1938: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1939: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1.77 noro 1940: nh = ndv_newps(0,nfv,nfqv);
1.20 noro 1941: d = update_pairs(d,g,nh);
1942: g = update_base(g,nh);
1943: } else {
1.71 noro 1944: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1945: }
1946: } else {
1.71 noro 1947: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1948: }
1949: FREENDP(l);
1950: }
1.77 noro 1951: if ( nd_demand )
1952: for ( t = g; t; t = NEXT(t) )
1953: BDY(t) = (pointer)ndv_load((int)BDY(t));
1954: else
1955: for ( t = g; t; t = NEXT(t) )
1956: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1957: return g;
1958: }
1959:
1.23 noro 1960: int ndv_compare(NDV *p1,NDV *p2)
1961: {
1.34 noro 1962: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1963: }
1964:
1965: int ndv_compare_rev(NDV *p1,NDV *p2)
1966: {
1.34 noro 1967: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1968: }
1969:
1.61 noro 1970: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1971: {
1.69 noro 1972: int i,n,stat;
1973: ND nf,g,head;
1.23 noro 1974: NODE t,a0,a;
1.69 noro 1975: union oNDC dn;
1976: NDV *w;
1977: Q q;
1.23 noro 1978:
1.69 noro 1979: n = length(f);
1980: #if 0
1981: w = (NDV *)ALLOCA(n*sizeof(NDV));
1982: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1983: qsort(w,n,sizeof(NDV),
1984: (int (*)(const void *,const void *))ndv_compare);
1985: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1986: #endif
1987: ndv_setup(m,0,f);
1988: for ( i = 0; i < n; ) {
1989: g = ndvtond(m,nd_ps[i]);
1990: g = nd_separate_head(g,&head);
1991: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1992: if ( !stat )
1.69 noro 1993: nd_reconstruct(m,0,0);
1994: else {
1.71 noro 1995: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1996: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1997: nf = nd_add(m,head,nf);
1998: ndv_free(nd_ps[i]);
1.24 noro 1999: nd_removecont(m,nf);
1.69 noro 2000: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
2001: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 2002: i++;
1.23 noro 2003: }
2004: }
1.71 noro 2005: if ( DP_Print ) { printf("\n"); }
1.23 noro 2006: for ( a0 = 0, i = 0; i < n; i++ ) {
2007: NEXTNODE(a0,a);
1.69 noro 2008: BDY(a) = (pointer)nd_ps[i];
1.23 noro 2009: }
2010: NEXT(a) = 0;
2011: return a0;
2012: }
2013:
1.1 noro 2014: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
2015: {
2016: ND_pairs d1,nd,cur,head,prev,remove;
2017:
2018: if ( !g ) return d;
2019: d = crit_B(d,t);
2020: d1 = nd_newpairs(g,t);
2021: d1 = crit_M(d1);
2022: d1 = crit_F(d1);
1.55 noro 2023: if ( do_weyl )
2024: head = d1;
2025: else {
2026: prev = 0; cur = head = d1;
2027: while ( cur ) {
2028: if ( crit_2( cur->i1,cur->i2 ) ) {
2029: remove = cur;
2030: if ( !prev ) head = cur = NEXT(cur);
2031: else cur = NEXT(prev) = NEXT(cur);
2032: FREENDP(remove);
2033: } else {
2034: prev = cur; cur = NEXT(cur);
2035: }
1.1 noro 2036: }
2037: }
2038: if ( !d )
2039: return head;
2040: else {
2041: nd = d;
1.34 noro 2042: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 2043: NEXT(nd) = head;
2044: return d;
2045: }
2046: }
2047:
2048: ND_pairs nd_newpairs( NODE g, int t )
2049: {
2050: NODE h;
1.61 noro 2051: UINT *dl;
1.34 noro 2052: int ts,s;
1.1 noro 2053: ND_pairs r,r0;
2054:
1.20 noro 2055: dl = DL(nd_psh[t]);
1.34 noro 2056: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 2057: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
2058: NEXTND_pairs(r0,r);
2059: r->i1 = (int)BDY(h);
2060: r->i2 = t;
1.20 noro 2061: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 2062: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
2063: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 2064: }
2065: NEXT(r) = 0;
2066: return r0;
2067: }
2068:
2069: ND_pairs crit_B( ND_pairs d, int s )
2070: {
2071: ND_pairs cur,head,prev,remove;
1.61 noro 2072: UINT *t,*tl,*lcm;
1.1 noro 2073: int td,tdl;
2074:
2075: if ( !d ) return 0;
1.20 noro 2076: t = DL(nd_psh[s]);
1.1 noro 2077: prev = 0;
2078: head = cur = d;
1.61 noro 2079: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 2080: while ( cur ) {
2081: tl = cur->lcm;
2082: if ( ndl_reducible(tl,t)
1.20 noro 2083: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
2084: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 2085: remove = cur;
2086: if ( !prev ) {
2087: head = cur = NEXT(cur);
2088: } else {
2089: cur = NEXT(prev) = NEXT(cur);
2090: }
2091: FREENDP(remove);
2092: } else {
1.34 noro 2093: prev = cur; cur = NEXT(cur);
1.1 noro 2094: }
2095: }
2096: return head;
2097: }
2098:
2099: ND_pairs crit_M( ND_pairs d1 )
2100: {
2101: ND_pairs e,d2,d3,dd,p;
1.61 noro 2102: UINT *id,*jd;
1.1 noro 2103:
2104: for ( dd = 0, e = d1; e; e = d3 ) {
2105: if ( !(d2 = NEXT(e)) ) {
2106: NEXT(e) = dd;
2107: return e;
2108: }
1.34 noro 2109: id = LCM(e);
1.1 noro 2110: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2111: p = NEXT(d2);
2112: jd = LCM(d2);
2113: if ( ndl_equal(jd,id) )
2114: ;
2115: else if ( TD(jd) > TD(id) )
1.1 noro 2116: if ( ndl_reducible(jd,id) ) continue;
2117: else ;
1.34 noro 2118: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2119: NEXT(d2) = d3;
2120: d3 = d2;
2121: }
2122: NEXT(e) = dd;
2123: dd = e;
2124: continue;
2125: /**/
2126: delit: NEXT(d2) = d3;
2127: d3 = d2;
2128: for ( ; p; p = d2 ) {
2129: d2 = NEXT(p);
2130: NEXT(p) = d3;
2131: d3 = p;
2132: }
2133: FREENDP(e);
2134: }
2135: return dd;
2136: }
2137:
2138: ND_pairs crit_F( ND_pairs d1 )
2139: {
2140: ND_pairs rest, head,remove;
2141: ND_pairs last, p, r, w;
2142: int s;
2143:
2144: for ( head = last = 0, p = d1; NEXT(p); ) {
2145: r = w = equivalent_pairs(p,&rest);
1.14 noro 2146: s = SG(r);
1.1 noro 2147: w = NEXT(w);
2148: while ( w ) {
2149: if ( crit_2(w->i1,w->i2) ) {
2150: r = w;
2151: w = NEXT(w);
2152: while ( w ) {
2153: remove = w;
2154: w = NEXT(w);
2155: FREENDP(remove);
2156: }
2157: break;
1.14 noro 2158: } else if ( SG(w) < s ) {
1.1 noro 2159: FREENDP(r);
2160: r = w;
1.14 noro 2161: s = SG(r);
1.1 noro 2162: w = NEXT(w);
2163: } else {
2164: remove = w;
2165: w = NEXT(w);
2166: FREENDP(remove);
2167: }
2168: }
2169: if ( last ) NEXT(last) = r;
2170: else head = r;
2171: NEXT(last = r) = 0;
2172: p = rest;
2173: if ( !p ) return head;
2174: }
2175: if ( !last ) return p;
2176: NEXT(last) = p;
2177: return head;
2178: }
2179:
2180: int crit_2( int dp1, int dp2 )
2181: {
1.20 noro 2182: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2183: }
2184:
1.40 noro 2185: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2186: {
2187: ND_pairs w,p,r,s;
1.61 noro 2188: UINT *d;
1.1 noro 2189:
2190: w = d1;
1.34 noro 2191: d = LCM(w);
1.1 noro 2192: s = NEXT(w);
2193: NEXT(w) = 0;
2194: for ( r = 0; s; s = p ) {
2195: p = NEXT(s);
1.34 noro 2196: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2197: NEXT(s) = w; w = s;
1.1 noro 2198: } else {
1.39 noro 2199: NEXT(s) = r; r = s;
1.1 noro 2200: }
2201: }
2202: *prest = r;
2203: return w;
2204: }
2205:
2206: NODE update_base(NODE nd,int ndp)
2207: {
1.61 noro 2208: UINT *dl, *dln;
1.1 noro 2209: NODE last, p, head;
2210:
1.20 noro 2211: dl = DL(nd_psh[ndp]);
1.1 noro 2212: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2213: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2214: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2215: p = NEXT(p);
2216: if ( last ) NEXT(last) = p;
2217: } else {
2218: if ( !last ) head = p;
2219: p = NEXT(last = p);
2220: }
2221: }
2222: head = append_one(head,ndp);
2223: return head;
2224: }
2225:
2226: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2227: {
2228: ND_pairs m,ml,p,l;
1.61 noro 2229: UINT *lcm;
1.33 noro 2230: int s,td,len,tlen,c,c1;
1.1 noro 2231:
2232: if ( !(p = NEXT(m = d)) ) {
2233: *prest = p;
2234: NEXT(m) = 0;
2235: return m;
2236: }
1.14 noro 2237: s = SG(m);
1.77 noro 2238: if ( !NoSugar ) {
2239: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2240: if ( (SG(p) < s)
2241: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
2242: ml = l; m = p; s = SG(m);
2243: }
2244: } else {
2245: for ( ml = 0, l = m; p; p = NEXT(l = p) )
2246: if ( DL_COMPARE(LCM(p),LCM(m)) < 0 ) {
2247: ml = l; m = p; s = SG(m);
2248: }
2249: }
1.1 noro 2250: if ( !ml ) *prest = NEXT(m);
2251: else {
2252: NEXT(ml) = NEXT(m);
2253: *prest = d;
2254: }
2255: NEXT(m) = 0;
2256: return m;
2257: }
2258:
1.63 noro 2259: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2260: {
1.69 noro 2261: int msugar,i;
1.63 noro 2262: ND_pairs t,dm0,dm,dr0,dr;
2263:
2264: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2265: if ( SG(t) < msugar ) msugar = SG(t);
2266: dm0 = 0; dr0 = 0;
1.69 noro 2267: for ( i = 0, t = d; t; t = NEXT(t) )
2268: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2269: if ( dm0 ) NEXT(dm) = t;
2270: else dm0 = t;
2271: dm = t;
1.69 noro 2272: i++;
1.63 noro 2273: } else {
2274: if ( dr0 ) NEXT(dr) = t;
2275: else dr0 = t;
2276: dr = t;
2277: }
2278: NEXT(dm) = 0;
2279: if ( dr0 ) NEXT(dr) = 0;
2280: *prest = dr0;
2281: return dm0;
2282: }
2283:
1.77 noro 2284: int ndv_newps(int m,NDV a,NDV aq)
1.1 noro 2285: {
1.3 noro 2286: int len;
1.13 noro 2287: RHist r;
1.20 noro 2288: NDV b;
1.3 noro 2289:
1.1 noro 2290: if ( nd_psn == nd_pslen ) {
2291: nd_pslen *= 2;
1.11 noro 2292: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2293: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2294: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2295: nd_bound = (UINT **)
2296: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2297: }
1.39 noro 2298: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2299: nd_ps[nd_psn] = a;
1.39 noro 2300: if ( aq ) {
1.63 noro 2301: nd_ps_trace[nd_psn] = aq;
2302: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2303: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2304: } else {
1.63 noro 2305: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2306: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2307: }
1.77 noro 2308: if ( nd_demand ) {
2309: if ( aq ) {
2310: ndv_save(nd_ps_trace[nd_psn],nd_psn);
2311: nd_ps_trace[nd_psn] = 0;
2312: } else {
2313: ndv_save(nd_ps[nd_psn],nd_psn);
2314: nd_ps[nd_psn] = 0;
2315: }
2316: }
1.1 noro 2317: return nd_psn++;
2318: }
2319:
1.61 noro 2320: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2321: {
1.5 noro 2322: int i,j,td,len,max;
1.1 noro 2323: NODE s,s0,f0;
1.61 noro 2324: UINT *d;
1.13 noro 2325: RHist r;
1.69 noro 2326: NDV *w;
1.61 noro 2327: NDV a,am;
1.11 noro 2328:
2329: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2330:
1.69 noro 2331: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2332: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2333: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2334: qsort(w,nd_psn,sizeof(NDV),
2335: (int (*)(const void *,const void *))ndv_compare);
2336: nd_pslen = 2*nd_psn;
1.11 noro 2337: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2338: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2339: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2340: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2341:
1.11 noro 2342: if ( !nd_red )
1.13 noro 2343: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2344: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2345: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2346: if ( trace ) {
1.69 noro 2347: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2348: ndv_removecont(0,a);
2349: am = nd_ps[i] = ndv_dup(mod,a);
2350: ndv_mod(mod,am);
2351: ndv_removecont(mod,am);
2352: } else {
1.69 noro 2353: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2354: ndv_removecont(mod,a);
1.39 noro 2355: }
1.61 noro 2356: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2357: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2358: nd_psh[i] = r;
1.77 noro 2359: if ( nd_demand ) {
2360: if ( trace ) {
2361: ndv_save(nd_ps_trace[i],i);
2362: nd_ps_trace[i] = 0;
2363: } else {
2364: ndv_save(nd_ps[i],i);
2365: nd_ps[i] = 0;
2366: }
2367: }
1.20 noro 2368: }
2369: }
2370:
1.63 noro 2371: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2372: {
1.61 noro 2373: VL tv,fv,vv,vc;
1.1 noro 2374: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2375: int e,max,nvar;
1.69 noro 2376: NDV b;
1.77 noro 2377: int ishomo;
1.1 noro 2378:
1.78 noro 2379: if ( !m && Demand ) nd_demand = 1;
2380: else nd_demand = 0;
2381:
1.74 noro 2382: ndv_alloc = 0;
1.1 noro 2383: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2384: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.32 noro 2385: nd_init_ord(ord);
1.61 noro 2386: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2387: for ( tv = vv; tv; tv = NEXT(tv) ) {
2388: e = getdeg(tv->v,(P)BDY(t));
2389: max = MAX(e,max);
2390: }
2391: nd_setup_parameters(nvar,max);
1.77 noro 2392: ishomo = 1;
1.1 noro 2393: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2394: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.77 noro 2395: if ( ishomo )
2396: ishomo = ishomo && ndv_ishomo(b);
1.69 noro 2397: if ( m ) ndv_mod(m,b);
1.61 noro 2398: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2399: }
2400: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2401: ndv_setup(m,0,fd0);
1.77 noro 2402: x = f4?nd_f4(m):nd_gb(m,ishomo,0);
1.82 ! noro 2403: nd_demand = 0;
1.61 noro 2404: x = ndv_reducebase(x);
2405: x = ndv_reduceall(m,x);
1.23 noro 2406: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2407: NEXTNODE(r0,r);
1.61 noro 2408: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2409: }
2410: if ( r0 ) NEXT(r) = 0;
2411: MKLIST(*rp,r0);
1.74 noro 2412: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2413: }
2414:
1.52 noro 2415: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2416: {
2417: struct order_spec ord1;
1.61 noro 2418: VL tv,fv,vv,vc;
1.27 noro 2419: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2420: int m,nocheck,nvar,mindex,e,max;
2421: NDV c;
2422: NMV a;
1.27 noro 2423: P p;
1.61 noro 2424: EPOS oepos;
1.82 ! noro 2425: int obpe,oadv,wmax,i,len,cbpe,ishomo;
1.20 noro 2426:
2427: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2428: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.52 noro 2429: nocheck = 0;
2430: mindex = 0;
2431:
1.78 noro 2432: if ( Demand ) nd_demand = 1;
2433: else nd_demand = 0;
2434:
1.52 noro 2435: /* setup modulus */
2436: if ( trace < 0 ) {
2437: trace = -trace;
2438: nocheck = 1;
2439: }
2440: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2441: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2442: for ( tv = vv; tv; tv = NEXT(tv) ) {
2443: e = getdeg(tv->v,(P)BDY(t));
2444: max = MAX(e,max);
1.23 noro 2445: }
1.61 noro 2446: nd_init_ord(ord);
2447: nd_setup_parameters(nvar,max);
2448: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
1.82 ! noro 2449: ishomo = 1;
1.61 noro 2450: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2451: c = ptondv(CO,vv,(P)BDY(t));
1.82 ! noro 2452: if ( ishomo )
! 2453: ishomo = ishomo && ndv_ishomo(c);
1.61 noro 2454: if ( c ) {
2455: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2456: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2457: }
1.61 noro 2458: }
2459: if ( in0 ) NEXT(in) = 0;
2460: if ( fd0 ) NEXT(fd) = 0;
1.82 ! noro 2461: if ( !ishomo && homo ) {
1.61 noro 2462: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2463: c = (NDV)BDY(t); len = LEN(c);
2464: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2465: wmax = MAX(TD(DL(a)),wmax);
2466: }
2467: homogenize_order(ord,nvar,&ord1);
2468: nd_init_ord(&ord1);
2469: nd_setup_parameters(nvar+1,wmax);
2470: for ( t = fd0; t; t = NEXT(t) )
2471: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2472: }
1.52 noro 2473: while ( 1 ) {
1.77 noro 2474: if ( Demand )
2475: nd_demand = 1;
1.61 noro 2476: ndv_setup(m,1,fd0);
1.82 ! noro 2477: cand = nd_gb_trace(m,ishomo || homo);
1.52 noro 2478: if ( !cand ) {
2479: /* failure */
1.61 noro 2480: if ( trace > 1 ) { *rp = 0; return; }
2481: else m = get_lprime(++mindex);
1.52 noro 2482: continue;
2483: }
1.82 ! noro 2484: if ( !ishomo && homo ) {
1.27 noro 2485: /* dehomogenization */
1.61 noro 2486: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2487: nd_init_ord(ord);
1.61 noro 2488: nd_setup_parameters(nvar,0);
1.27 noro 2489: }
1.77 noro 2490: nd_demand = 0;
1.61 noro 2491: cand = ndv_reducebase(cand);
2492: cand = ndv_reduceall(0,cand);
1.78 noro 2493: cbpe = nd_bpe;
1.61 noro 2494: if ( nocheck )
2495: break;
2496: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2497: /* success */
2498: break;
2499: else if ( trace > 1 ) {
2500: /* failure */
2501: *rp = 0; return;
1.61 noro 2502: } else {
1.52 noro 2503: /* try the next modulus */
2504: m = get_lprime(++mindex);
1.61 noro 2505: /* reset the parameters */
1.82 ! noro 2506: if ( !ishomo && homo ) {
1.61 noro 2507: nd_init_ord(&ord1);
2508: nd_setup_parameters(nvar+1,wmax);
2509: } else {
2510: nd_init_ord(ord);
2511: nd_setup_parameters(nvar,max);
2512: }
2513: }
1.52 noro 2514: }
1.27 noro 2515: /* dp->p */
1.61 noro 2516: nd_bpe = cbpe;
1.73 noro 2517: nd_setup_parameters(nd_nvar,0);
1.61 noro 2518: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2519: MKLIST(*rp,cand);
1.1 noro 2520: }
2521:
1.61 noro 2522: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2523: {
1.61 noro 2524: UINT *d;
1.57 noro 2525: int i,j,l,s,ord_l;
1.43 noro 2526: struct order_pair *op;
1.1 noro 2527:
2528: d = dl->d;
1.41 noro 2529: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2530: if ( nd_blockmask ) {
2531: l = nd_blockmask->n;
2532: op = nd_blockmask->order_pair;
2533: for ( j = 0, s = 0; j < l; j++ ) {
2534: ord_l = op[j].length;
1.57 noro 2535: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2536: }
2537: TD(r) = ndl_weight(r);
1.61 noro 2538: ndl_weight_mask(r);
1.43 noro 2539: } else {
1.56 noro 2540: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2541: TD(r) = ndl_weight(r);
2542: }
1.1 noro 2543: }
2544:
1.61 noro 2545: DL ndltodl(int n,UINT *ndl)
1.1 noro 2546: {
2547: DL dl;
2548: int *d;
1.57 noro 2549: int i,j,l,s,ord_l;
1.43 noro 2550: struct order_pair *op;
1.1 noro 2551:
2552: NEWDL(dl,n);
1.34 noro 2553: dl->td = TD(ndl);
1.1 noro 2554: d = dl->d;
1.43 noro 2555: if ( nd_blockmask ) {
2556: l = nd_blockmask->n;
2557: op = nd_blockmask->order_pair;
2558: for ( j = 0, s = 0; j < l; j++ ) {
2559: ord_l = op[j].length;
1.57 noro 2560: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2561: }
2562: } else {
1.56 noro 2563: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2564: }
1.1 noro 2565: return dl;
2566: }
2567:
1.61 noro 2568: void ndl_print(UINT *dl)
1.1 noro 2569: {
2570: int n;
1.57 noro 2571: int i,j,l,ord_l,s,s0;
1.43 noro 2572: struct order_pair *op;
1.1 noro 2573:
2574: n = nd_nvar;
2575: printf("<<");
1.43 noro 2576: if ( nd_blockmask ) {
2577: l = nd_blockmask->n;
2578: op = nd_blockmask->order_pair;
2579: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2580: ord_l = op[j].length;
1.57 noro 2581: for ( i = 0; i < ord_l; i++, s++ )
2582: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2583: }
2584: } else {
1.56 noro 2585: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2586: }
1.1 noro 2587: printf(">>");
2588: }
2589:
2590: void nd_print(ND p)
2591: {
2592: NM m;
2593:
2594: if ( !p )
2595: printf("0\n");
2596: else {
2597: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2598: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2599: else printf("+%d*",CM(m));
1.14 noro 2600: ndl_print(DL(m));
1.1 noro 2601: }
2602: printf("\n");
2603: }
2604: }
2605:
1.16 noro 2606: void nd_print_q(ND p)
2607: {
2608: NM m;
2609:
2610: if ( !p )
2611: printf("0\n");
2612: else {
2613: for ( m = BDY(p); m; m = NEXT(m) ) {
2614: printf("+");
1.74 noro 2615: printexpr(CO,(Obj)CQ(m));
1.16 noro 2616: printf("*");
2617: ndl_print(DL(m));
2618: }
2619: printf("\n");
2620: }
2621: }
2622:
1.1 noro 2623: void ndp_print(ND_pairs d)
2624: {
2625: ND_pairs t;
2626:
1.34 noro 2627: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2628: printf("\n");
2629: }
2630:
1.20 noro 2631: void nd_removecont(int mod,ND p)
1.16 noro 2632: {
2633: int i,n;
2634: Q *w;
2635: Q dvr,t;
2636: NM m;
1.21 noro 2637: struct oVECT v;
2638: N q,r;
1.16 noro 2639:
1.71 noro 2640: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2641: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2642: else {
2643: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2644: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2645: v.len = n;
2646: v.body = (pointer *)w;
1.34 noro 2647: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2648: removecont_array(w,n);
2649: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2650: }
2651: }
2652:
1.21 noro 2653: void nd_removecont2(ND p1,ND p2)
2654: {
2655: int i,n1,n2,n;
2656: Q *w;
2657: Q dvr,t;
2658: NM m;
2659: struct oVECT v;
2660: N q,r;
2661:
2662: if ( !p1 ) {
2663: nd_removecont(0,p2); return;
2664: } else if ( !p2 ) {
2665: nd_removecont(0,p1); return;
2666: }
2667: n1 = nd_length(p1);
2668: n2 = nd_length(p2);
2669: n = n1+n2;
2670: w = (Q *)ALLOCA(n*sizeof(Q));
2671: v.len = n;
2672: v.body = (pointer *)w;
1.34 noro 2673: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2674: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2675: removecont_array(w,n);
2676: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2677: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2678: }
2679:
1.20 noro 2680: void ndv_removecont(int mod,NDV p)
1.16 noro 2681: {
2682: int i,len;
2683: Q *w;
2684: Q dvr,t;
2685: NMV m;
2686:
1.71 noro 2687: if ( mod == -1 )
2688: ndv_mul_c(mod,p,_invsf(HCM(p)));
2689: else if ( mod )
1.20 noro 2690: ndv_mul_c(mod,p,invm(HCM(p),mod));
2691: else {
2692: len = p->len;
2693: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2694: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2695: sortbynm(w,len);
2696: qltozl(w,len,&dvr);
2697: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2698: divq(CQ(m),dvr,&t); CQ(m) = t;
2699: }
1.16 noro 2700: }
1.21 noro 2701: }
2702:
1.61 noro 2703: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2704: {
2705: int len,i,max;
2706: NMV m,mr0,mr,t;
2707:
2708: len = p->len;
2709: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2710: max = MAX(max,TD(DL(m)));
2711: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2712: m = (NMV)((char *)mr0+(len-1)*oadv);
2713: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2714: t = (NMV)ALLOCA(nmv_adv);
2715: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2716: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2717: CQ(mr) = CQ(m);
2718: ndl_copy(DL(t),DL(mr));
2719: }
2720: NV(p)++;
2721: BDY(p) = mr0;
2722: }
2723:
1.45 noro 2724: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2725: {
1.45 noro 2726: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2727: Q *w;
2728: Q dvr,t;
2729: NMV m,r;
2730:
2731: len = p->len;
2732: newnvar = nd_nvar-1;
1.48 noro 2733: newexporigin = nd_get_exporigin(ord);
1.45 noro 2734: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2735: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2736: ndl_dehomogenize(DL(m));
1.23 noro 2737: if ( newwpd != nd_wpd ) {
1.61 noro 2738: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2739: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2740: CQ(r) = CQ(m);
2741: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2742: adj = nd_exporigin-newexporigin;
2743: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2744: }
2745: }
2746: NV(p)--;
2747: }
2748:
1.21 noro 2749: void removecont_array(Q *c,int n)
2750: {
2751: struct oVECT v;
2752: Q d0,d1,a,u,u1,gcd;
2753: int i;
2754: N qn,rn,gn;
2755: Q *q,*r;
2756:
2757: q = (Q *)ALLOCA(n*sizeof(Q));
2758: r = (Q *)ALLOCA(n*sizeof(Q));
2759: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2760: igcdv_estimate(&v,&d0);
2761: for ( i = 0; i < n; i++ ) {
2762: divn(NM(c[i]),NM(d0),&qn,&rn);
2763: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2764: NTOQ(rn,SGN(c[i]),r[i]);
2765: }
1.34 noro 2766: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2767: if ( i < n ) {
2768: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2769: igcdv(&v,&d1);
2770: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2771: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2772: for ( i = 0; i < n; i++ ) {
2773: mulq(a,q[i],&u);
2774: if ( r[i] ) {
2775: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2776: addq(u,u1,&q[i]);
2777: } else
2778: q[i] = u;
2779: }
2780: }
1.34 noro 2781: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2782: }
2783:
1.19 noro 2784: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2785: {
2786: NM m;
2787: int c,c1;
2788:
1.34 noro 2789: if ( !p ) return;
1.71 noro 2790: if ( mod == -1 )
2791: for ( m = BDY(p); m; m = NEXT(m) )
2792: CM(m) = _mulsf(CM(m),mul);
2793: else
2794: for ( m = BDY(p); m; m = NEXT(m) ) {
2795: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2796: }
1.1 noro 2797: }
2798:
1.16 noro 2799: void nd_mul_c_q(ND p,Q mul)
2800: {
2801: NM m;
2802: Q c;
2803:
1.34 noro 2804: if ( !p ) return;
1.16 noro 2805: for ( m = BDY(p); m; m = NEXT(m) ) {
2806: mulq(CQ(m),mul,&c); CQ(m) = c;
2807: }
2808: }
2809:
1.61 noro 2810: void nd_mul_c_p(VL vl,ND p,P mul)
2811: {
2812: NM m;
2813: P c;
2814:
2815: if ( !p ) return;
2816: for ( m = BDY(p); m; m = NEXT(m) ) {
2817: mulp(vl,CP(m),mul,&c); CP(m) = c;
2818: }
2819: }
2820:
1.1 noro 2821: void nd_free(ND p)
2822: {
2823: NM t,s;
2824:
1.34 noro 2825: if ( !p ) return;
1.1 noro 2826: t = BDY(p);
2827: while ( t ) {
2828: s = NEXT(t);
2829: FREENM(t);
2830: t = s;
2831: }
2832: FREEND(p);
2833: }
2834:
1.23 noro 2835: void ndv_free(NDV p)
2836: {
2837: GC_free(BDY(p));
2838: }
2839:
1.61 noro 2840: void nd_append_red(UINT *d,int i)
1.1 noro 2841: {
1.13 noro 2842: RHist m,m0;
1.1 noro 2843: int h;
2844:
1.13 noro 2845: NEWRHist(m);
1.34 noro 2846: h = ndl_hash_value(d);
1.13 noro 2847: m->index = i;
1.14 noro 2848: ndl_copy(d,DL(m));
1.1 noro 2849: NEXT(m) = nd_red[h];
2850: nd_red[h] = m;
2851: }
2852:
1.61 noro 2853: UINT *ndv_compute_bound(NDV p)
1.1 noro 2854: {
1.61 noro 2855: UINT *d1,*d2,*t;
2856: UINT u;
1.57 noro 2857: int i,j,k,l,len,ind;
1.45 noro 2858: NMV m;
1.1 noro 2859:
2860: if ( !p )
2861: return 0;
1.61 noro 2862: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2863: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2864: len = LEN(p);
2865: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2866: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2867: ndl_lcm(DL(m),d1,d2);
1.1 noro 2868: t = d1; d1 = d2; d2 = t;
2869: }
1.12 noro 2870: l = nd_nvar+31;
1.61 noro 2871: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2872: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2873: u = d1[i];
2874: k = (nd_epw-1)*nd_bpe;
2875: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2876: t[ind] = (u>>k)&nd_mask0;
2877: }
2878: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2879: return t;
2880: }
2881:
1.48 noro 2882: int nd_get_exporigin(struct order_spec *ord)
2883: {
1.51 noro 2884: switch ( ord->id ) {
1.41 noro 2885: case 0:
1.48 noro 2886: return 1;
1.41 noro 2887: case 1:
2888: /* block order */
1.43 noro 2889: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2890: return ord->ord.block.length+1;
1.41 noro 2891: case 2:
1.52 noro 2892: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2893: }
1.48 noro 2894: }
2895:
1.61 noro 2896: void nd_setup_parameters(int nvar,int max) {
1.79 noro 2897: int i,j,n,elen,ord_o,ord_l,l,s,wpd;
1.57 noro 2898: struct order_pair *op;
1.48 noro 2899:
1.73 noro 2900: nd_nvar = nvar;
1.79 noro 2901: if ( max ) {
2902: if ( max < 2 ) nd_bpe = 1;
2903: else if ( max < 4 ) nd_bpe = 2;
2904: else if ( max < 8 ) nd_bpe = 3;
2905: else if ( max < 16 ) nd_bpe = 4;
2906: else if ( max < 32 ) nd_bpe = 5;
2907: else if ( max < 64 ) nd_bpe = 6;
2908: else if ( max < 256 ) nd_bpe = 8;
2909: else if ( max < 1024 ) nd_bpe = 10;
2910: else if ( max < 65536 ) nd_bpe = 16;
2911: else nd_bpe = 32;
2912: }
1.61 noro 2913: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2914: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2915: nd_exporigin = nd_get_exporigin(nd_ord);
1.79 noro 2916: wpd = nd_exporigin+elen;
2917: if ( wpd != nd_wpd ) {
2918: nd_free_private_storage();
2919: nd_wpd = wpd;
2920: }
1.1 noro 2921: if ( nd_bpe < 32 ) {
2922: nd_mask0 = (1<<nd_bpe)-1;
2923: } else {
2924: nd_mask0 = 0xffffffff;
2925: }
2926: bzero(nd_mask,sizeof(nd_mask));
2927: nd_mask1 = 0;
2928: for ( i = 0; i < nd_epw; i++ ) {
2929: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2930: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2931: }
1.61 noro 2932: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2933: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2934: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2935: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2936: }
2937:
1.20 noro 2938: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2939: {
1.37 noro 2940: int i,obpe,oadv,h;
1.71 noro 2941: static NM prev_nm_free_list;
2942: static ND_pairs prev_ndp_free_list;
1.13 noro 2943: RHist mr0,mr;
2944: RHist r;
1.37 noro 2945: RHist *old_red;
1.71 noro 2946: ND_pairs s0,s,t;
1.43 noro 2947: EPOS oepos;
1.15 noro 2948:
1.1 noro 2949: obpe = nd_bpe;
1.11 noro 2950: oadv = nmv_adv;
1.43 noro 2951: oepos = nd_epos;
1.63 noro 2952: if ( obpe < 2 ) nd_bpe = 2;
2953: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2954: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2955: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2956: else if ( obpe < 6 ) nd_bpe = 6;
2957: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2958: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2959: else if ( obpe < 16 ) nd_bpe = 16;
2960: else if ( obpe < 32 ) nd_bpe = 32;
2961: else error("nd_reconstruct : exponent too large");
1.5 noro 2962:
1.73 noro 2963: nd_setup_parameters(nd_nvar,0);
1.1 noro 2964: prev_nm_free_list = _nm_free_list;
2965: prev_ndp_free_list = _ndp_free_list;
2966: _nm_free_list = 0;
2967: _ndp_free_list = 0;
1.53 noro 2968: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2969: if ( trace )
2970: for ( i = nd_psn-1; i >= 0; i-- )
2971: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2972: s0 = 0;
2973: for ( t = d; t; t = NEXT(t) ) {
2974: NEXTND_pairs(s0,s);
2975: s->i1 = t->i1;
2976: s->i2 = t->i2;
1.14 noro 2977: SG(s) = SG(t);
1.61 noro 2978: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2979: }
1.37 noro 2980:
2981: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2982: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2983: old_red[i] = nd_red[i];
2984: nd_red[i] = 0;
2985: }
2986: for ( i = 0; i < REDTAB_LEN; i++ )
2987: for ( r = old_red[i]; r; r = NEXT(r) ) {
2988: NEWRHist(mr);
1.13 noro 2989: mr->index = r->index;
1.20 noro 2990: SG(mr) = SG(r);
1.61 noro 2991: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2992: h = ndl_hash_value(DL(mr));
2993: NEXT(mr) = nd_red[h];
2994: nd_red[h] = mr;
1.6 noro 2995: }
1.37 noro 2996: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2997: old_red = 0;
1.11 noro 2998: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2999: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 3000: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 3001: nd_psh[i] = r;
1.11 noro 3002: }
1.1 noro 3003: if ( s0 ) NEXT(s) = 0;
3004: prev_nm_free_list = 0;
3005: prev_ndp_free_list = 0;
1.71 noro 3006: #if 0
1.1 noro 3007: GC_gcollect();
1.71 noro 3008: #endif
1.1 noro 3009: return s0;
3010: }
3011:
1.61 noro 3012: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 3013: {
1.57 noro 3014: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 3015: struct order_pair *op;
1.1 noro 3016:
3017: n = nd_nvar;
1.61 noro 3018: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 3019: omask0 = (1<<obpe)-1;
1.34 noro 3020: TD(r) = TD(d);
1.41 noro 3021: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 3022: if ( nd_blockmask ) {
3023: l = nd_blockmask->n;
3024: op = nd_blockmask->order_pair;
3025: for ( i = 1; i < nd_exporigin; i++ )
3026: r[i] = d[i];
3027: for ( j = 0, s = 0; j < l; j++ ) {
3028: ord_l = op[j].length;
1.57 noro 3029: for ( i = 0; i < ord_l; i++, s++ ) {
3030: ei = GET_EXP_OLD(d,s);
3031: PUT_EXP(r,s,ei);
3032: }
1.1 noro 3033: }
1.43 noro 3034: } else {
1.56 noro 3035: for ( i = 0; i < n; i++ ) {
3036: ei = GET_EXP_OLD(d,i);
3037: PUT_EXP(r,i,ei);
3038: }
1.1 noro 3039: }
3040: }
1.3 noro 3041:
1.6 noro 3042: ND nd_copy(ND p)
3043: {
3044: NM m,mr,mr0;
1.41 noro 3045: int c,n;
1.6 noro 3046: ND r;
3047:
3048: if ( !p )
3049: return 0;
3050: else {
3051: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
3052: NEXTNM(mr0,mr);
1.14 noro 3053: CM(mr) = CM(m);
3054: ndl_copy(DL(m),DL(mr));
1.6 noro 3055: }
3056: NEXT(mr) = 0;
1.31 noro 3057: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 3058: SG(r) = SG(p);
1.6 noro 3059: return r;
3060: }
3061: }
3062:
1.53 noro 3063: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 3064: {
3065: NM m;
3066: NDV p1,p2;
3067: ND t1,t2;
1.61 noro 3068: UINT *lcm;
1.31 noro 3069: int td;
1.11 noro 3070:
1.77 noro 3071: if ( !mod && nd_demand ) {
3072: p1 = ndv_load(p->i1); p2 = ndv_load(p->i2);
1.53 noro 3073: } else {
1.77 noro 3074: if ( trace ) {
3075: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
3076: } else {
3077: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
3078: }
1.20 noro 3079: }
1.34 noro 3080: lcm = LCM(p);
1.11 noro 3081: NEWNM(m);
1.20 noro 3082: CQ(m) = HCQ(p2);
1.34 noro 3083: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 3084: if ( ndl_check_bound2(p->i1,DL(m)) )
3085: return 0;
1.55 noro 3086: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 3087: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
3088: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 3089: else chsgnq(HCQ(p1),&CQ(m));
3090: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 3091: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 3092: nd_free(t1);
3093: return 0;
3094: }
1.55 noro 3095: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 3096: *rp = nd_add(mod,t1,t2);
1.11 noro 3097: FREENM(m);
3098: return 1;
3099: }
3100:
1.19 noro 3101: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 3102: {
3103: NMV m;
3104: int c,c1,len,i;
3105:
1.34 noro 3106: if ( !p ) return;
1.14 noro 3107: len = LEN(p);
1.71 noro 3108: if ( mod == -1 )
3109: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
3110: CM(m) = _mulsf(CM(m),mul);
3111: else
3112: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3113: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
3114: }
1.11 noro 3115: }
3116:
1.16 noro 3117: void ndv_mul_c_q(NDV p,Q mul)
3118: {
3119: NMV m;
3120: Q c;
3121: int len,i;
3122:
1.34 noro 3123: if ( !p ) return;
1.16 noro 3124: len = LEN(p);
3125: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3126: mulq(CQ(m),mul,&c); CQ(m) = c;
3127: }
3128: }
3129:
1.55 noro 3130: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
3131: int n2,i,j,l,n,tlen;
1.61 noro 3132: UINT *d0;
1.55 noro 3133: NM *tab,*psum;
3134: ND s,r;
3135: NM t;
3136: NMV m1;
3137:
3138: if ( !p ) return 0;
3139: n = NV(p); n2 = n>>1;
3140: d0 = DL(m0);
3141: l = LEN(p);
3142: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
3143: tab = (NM *)ALLOCA(tlen*sizeof(NM));
3144: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3145: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3146: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3147: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3148: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3149: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3150: for ( j = 0; j < tlen; j++ ) {
3151: if ( tab[j] ) {
3152: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3153: }
3154: }
3155: }
3156: for ( i = tlen-1, r = 0; i >= 0; i-- )
3157: if ( psum[i] ) {
3158: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3159: MKND(n,psum[i],j,s);
3160: r = nd_add(mod,r,s);
3161: }
1.56 noro 3162: if ( r ) SG(r) = SG(p)+TD(d0);
3163: return r;
1.55 noro 3164: }
3165:
1.56 noro 3166: /* product of monomials */
3167: /* XXX block order is not handled correctly */
3168:
1.55 noro 3169: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3170: {
1.56 noro 3171: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3172: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3173: Q *ctab_q;
3174: Q q,q1;
1.61 noro 3175: UINT c0,c1,c;
1.55 noro 3176: NM *p;
3177: NM m,t;
3178:
3179: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3180: if ( !m0 || !m1 ) return;
3181: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3182: NEWNM(m); d = DL(m);
1.56 noro 3183: if ( mod ) {
3184: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3185: } else
3186: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3187: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3188: homo = n&1 ? 1 : 0;
3189: if ( homo ) {
3190: /* offset of h-degree */
3191: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3192: PUT_EXP(DL(m),n-1,h);
3193: TD(DL(m)) = h;
1.61 noro 3194: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3195: }
3196: tab[0] = m;
3197: NEWNM(m); d = DL(m);
1.57 noro 3198: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3199: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3200: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3201: /* xi^a*(Di^k*xi^l)*Di^b */
3202: a += l; b += k;
1.56 noro 3203: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3204: if ( !k || !l ) {
3205: for ( j = 0; j < curlen; j++ )
1.56 noro 3206: if ( t = tab[j] ) {
3207: dt = DL(t);
3208: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3209: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3210: }
3211: curlen *= k+1;
3212: continue;
3213: }
3214: min = MIN(k,l);
1.56 noro 3215: if ( mod ) {
1.61 noro 3216: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3217: mkwcm(k,l,mod,ctab);
3218: } else {
3219: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3220: mkwc(k,l,ctab_q);
3221: }
1.57 noro 3222: for ( j = min; j >= 0; j-- ) {
1.56 noro 3223: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3224: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3225: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3226: if ( homo ) {
3227: TD(d) = s;
1.56 noro 3228: PUT_EXP(d,n-1,s-h);
1.55 noro 3229: } else TD(d) = h;
1.61 noro 3230: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3231: if ( mod ) c = ctab[j];
3232: else q = ctab_q[j];
1.57 noro 3233: p = tab+curlen*j;
3234: if ( j == 0 ) {
3235: for ( u = 0; u < curlen; u++, p++ ) {
3236: if ( tab[u] ) {
3237: ndl_addto(DL(tab[u]),d);
3238: if ( mod ) {
3239: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3240: } else {
3241: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3242: }
3243: }
1.56 noro 3244: }
1.57 noro 3245: } else {
3246: for ( u = 0; u < curlen; u++, p++ ) {
3247: if ( tab[u] ) {
3248: NEWNM(t);
3249: ndl_add(DL(tab[u]),d,DL(t));
3250: if ( mod ) {
3251: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3252: } else
3253: mulq(CQ(tab[u]),q,&CQ(t));
3254: *p = t;
3255: }
1.55 noro 3256: }
3257: }
3258: }
3259: curlen *= k+1;
3260: }
3261: FREENM(m);
3262: }
3263:
1.63 noro 3264: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3265: {
3266: NM mr,mr0;
3267: NMV m;
3268: UINT *d,*dt,*dm;
3269: int c,n,td,i,c1,c2,len;
3270: Q q;
3271: ND r;
3272:
3273: if ( !p ) return 0;
3274: else {
3275: n = NV(p); m = BDY(p);
3276: d = DL(m0);
3277: len = LEN(p);
3278: mr0 = 0;
3279: td = TD(d);
3280: c = CM(m0);
3281: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3282: NEXTNM(mr0,mr);
3283: CM(mr) = 1;
3284: ndl_add(DL(m),d,DL(mr));
3285: }
3286: NEXT(mr) = 0;
3287: MKND(NV(p),mr0,len,r);
3288: SG(r) = SG(p) + TD(d);
3289: return r;
3290: }
3291: }
3292:
1.55 noro 3293: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3294: {
3295: NM mr,mr0;
3296: NMV m;
1.61 noro 3297: UINT *d,*dt,*dm;
1.9 noro 3298: int c,n,td,i,c1,c2,len;
1.16 noro 3299: Q q;
1.9 noro 3300: ND r;
3301:
1.34 noro 3302: if ( !p ) return 0;
1.55 noro 3303: else if ( do_weyl )
1.71 noro 3304: if ( mod == -1 )
3305: error("ndv_mul_nm : not implemented (weyl)");
3306: else
3307: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3308: else {
3309: n = NV(p); m = BDY(p);
1.34 noro 3310: d = DL(m0);
1.14 noro 3311: len = LEN(p);
1.9 noro 3312: mr0 = 0;
1.34 noro 3313: td = TD(d);
1.71 noro 3314: if ( mod == -1 ) {
3315: c = CM(m0);
3316: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3317: NEXTNM(mr0,mr);
3318: CM(mr) = _mulsf(CM(m),c);
3319: ndl_add(DL(m),d,DL(mr));
3320: }
3321: } else if ( mod ) {
1.16 noro 3322: c = CM(m0);
3323: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3324: NEXTNM(mr0,mr);
3325: c1 = CM(m);
1.19 noro 3326: DMAR(c1,c,0,mod,c2);
1.16 noro 3327: CM(mr) = c2;
3328: ndl_add(DL(m),d,DL(mr));
3329: }
3330: } else {
3331: q = CQ(m0);
3332: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3333: NEXTNM(mr0,mr);
3334: mulq(CQ(m),q,&CQ(mr));
3335: ndl_add(DL(m),d,DL(mr));
3336: }
1.4 noro 3337: }
1.9 noro 3338: NEXT(mr) = 0;
1.31 noro 3339: MKND(NV(p),mr0,len,r);
1.34 noro 3340: SG(r) = SG(p) + TD(d);
1.9 noro 3341: return r;
1.4 noro 3342: }
3343: }
3344:
1.43 noro 3345: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3346: {
1.13 noro 3347: NMV m,mr,mr0,t;
3348: int len,i,k;
1.11 noro 3349:
1.61 noro 3350: if ( !p ) return;
3351: m = BDY(p); len = LEN(p);
3352: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3353: m = (NMV)((char *)mr0+(len-1)*oadv);
3354: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3355: t = (NMV)ALLOCA(nmv_adv);
3356: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3357: CQ(t) = CQ(m);
3358: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3359: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3360: CQ(mr) = CQ(t);
3361: ndl_copy(DL(t),DL(mr));
3362: }
3363: BDY(p) = mr0;
3364: }
3365:
3366: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3367: {
3368: NMV m,mr,mr0;
3369: int len,i;
3370: NDV r;
1.11 noro 3371:
1.61 noro 3372: if ( !p ) return 0;
3373: m = BDY(p); len = LEN(p);
3374: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3375: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3376: ndl_zero(DL(mr));
3377: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3378: CQ(mr) = CQ(m);
1.11 noro 3379: }
1.61 noro 3380: MKNDV(NV(p),mr0,len,r);
3381: SG(r) = SG(p);
3382: return r;
1.11 noro 3383: }
3384:
1.61 noro 3385: /* duplicate p */
3386:
3387: NDV ndv_dup(int mod,NDV p)
1.3 noro 3388: {
3389: NDV d;
1.61 noro 3390: NMV t,m,m0;
1.3 noro 3391: int i,len;
3392:
1.34 noro 3393: if ( !p ) return 0;
1.31 noro 3394: len = LEN(p);
1.34 noro 3395: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3396: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3397: ndl_copy(DL(t),DL(m));
1.16 noro 3398: CQ(m) = CQ(t);
1.3 noro 3399: }
3400: MKNDV(NV(p),m0,len,d);
1.23 noro 3401: SG(d) = SG(p);
3402: return d;
3403: }
3404:
1.63 noro 3405: ND nd_dup(ND p)
3406: {
3407: ND d;
3408: NM t,m,m0;
3409:
3410: if ( !p ) return 0;
3411: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3412: NEXTNM(m0,m);
3413: ndl_copy(DL(t),DL(m));
3414: CQ(m) = CQ(t);
3415: }
3416: if ( m0 ) NEXT(m) = 0;
3417: MKND(NV(p),m0,LEN(p),d);
3418: SG(d) = SG(p);
3419: return d;
3420: }
3421:
1.61 noro 3422: /* XXX if p->len == 0 then it represents 0 */
3423:
3424: void ndv_mod(int mod,NDV p)
3425: {
3426: NMV t,d;
3427: int r;
3428: int i,len,dlen;
1.71 noro 3429: Obj gfs;
1.61 noro 3430:
3431: if ( !p ) return;
3432: len = LEN(p);
3433: dlen = 0;
1.71 noro 3434: if ( mod == -1 )
3435: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3436: simp_ff((Obj)CP(t),&gfs);
3437: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3438: CM(d) = r;
3439: ndl_copy(DL(t),DL(d));
3440: NMV_ADV(d);
3441: dlen++;
3442: }
1.71 noro 3443: else
3444: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3445: r = rem(NM(CQ(t)),mod);
3446: if ( r ) {
3447: if ( SGN(CQ(t)) < 0 )
3448: r = mod-r;
3449: CM(d) = r;
3450: ndl_copy(DL(t),DL(d));
3451: NMV_ADV(d);
3452: dlen++;
3453: }
3454: }
1.61 noro 3455: LEN(p) = dlen;
3456: }
3457:
3458: NDV ptondv(VL vl,VL dvl,P p)
3459: {
3460: ND nd;
3461:
3462: nd = ptond(vl,dvl,p);
3463: return ndtondv(0,nd);
3464: }
3465:
3466: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3467: {
1.61 noro 3468: int n,i,j,k,e;
3469: VL tvl;
3470: V v;
3471: DCP dc;
3472: DCP *w;
3473: ND r,s,t,u;
3474: P x;
3475: int c;
3476: UINT *d;
1.23 noro 3477: NM m,m0;
1.61 noro 3478:
3479: if ( !p )
3480: return 0;
3481: else if ( NUM(p) ) {
3482: NEWNM(m);
3483: ndl_zero(DL(m));
3484: CQ(m) = (Q)p;
3485: NEXT(m) = 0;
3486: MKND(nd_nvar,m,1,r);
3487: SG(r) = 0;
3488: return r;
3489: } else {
3490: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3491: w = (DCP *)ALLOCA(k*sizeof(DCP));
3492: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3493: for ( i = 0, tvl = dvl, v = VR(p);
3494: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3495: if ( !tvl ) {
3496: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3497: t = ptond(vl,dvl,COEF(w[j]));
3498: pwrp(vl,x,DEG(w[j]),&p);
3499: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3500: }
3501: return s;
3502: } else {
3503: NEWNM(m0); d = DL(m0);
3504: for ( j = k-1, s = 0; j >= 0; j-- ) {
3505: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3506: TD(d) = MUL_WEIGHT(e,i);
3507: if ( nd_blockmask) ndl_weight_mask(d);
3508: t = ptond(vl,dvl,COEF(w[j]));
3509: for ( m = BDY(t); m; m = NEXT(m) )
3510: ndl_addto(DL(m),d);
3511: SG(t) += TD(d);
3512: s = nd_add(0,s,t);
3513: }
3514: FREENM(m0);
3515: return s;
3516: }
3517: }
3518: }
3519:
3520: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3521: {
3522: VL tvl;
3523: int len,n,j,i,e;
3524: NMV m;
3525: Q q;
3526: P c;
3527: UINT *d;
3528: P s,r,u,t,w;
1.71 noro 3529: GFS gfs;
1.23 noro 3530:
1.34 noro 3531: if ( !p ) return 0;
1.61 noro 3532: else {
3533: len = LEN(p);
3534: n = NV(p);
3535: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3536: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3537: if ( mod == -1 ) {
3538: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3539: } else if ( mod ) {
1.61 noro 3540: STOQ(CM(m),q); c = (P)q;
3541: } else
3542: c = CP(m);
3543: d = DL(m);
3544: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3545: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3546: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3547: }
3548: addp(vl,s,t,&u); s = u;
3549: }
3550: return s;
1.23 noro 3551: }
1.3 noro 3552: }
3553:
1.61 noro 3554: NDV ndtondv(int mod,ND p)
1.11 noro 3555: {
3556: NDV d;
1.61 noro 3557: NMV m,m0;
3558: NM t;
3559: int i,len;
1.11 noro 3560:
1.34 noro 3561: if ( !p ) return 0;
1.61 noro 3562: len = LEN(p);
1.74 noro 3563: if ( mod )
3564: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3565: else
3566: m0 = m = MALLOC(len*nmv_adv);
3567: ndv_alloc += nmv_adv*len;
1.61 noro 3568: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3569: ndl_copy(DL(t),DL(m));
3570: CQ(m) = CQ(t);
1.11 noro 3571: }
1.61 noro 3572: MKNDV(NV(p),m0,len,d);
1.14 noro 3573: SG(d) = SG(p);
1.11 noro 3574: return d;
3575: }
3576:
1.61 noro 3577: ND ndvtond(int mod,NDV p)
1.11 noro 3578: {
1.61 noro 3579: ND d;
3580: NM m,m0;
1.11 noro 3581: NMV t;
1.61 noro 3582: int i,len;
1.11 noro 3583:
1.34 noro 3584: if ( !p ) return 0;
1.11 noro 3585: m0 = 0;
1.61 noro 3586: len = p->len;
3587: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3588: NEXTNM(m0,m);
3589: ndl_copy(DL(t),DL(m));
3590: CQ(m) = CQ(t);
1.11 noro 3591: }
3592: NEXT(m) = 0;
1.61 noro 3593: MKND(NV(p),m0,len,d);
1.14 noro 3594: SG(d) = SG(p);
1.11 noro 3595: return d;
3596: }
3597:
1.3 noro 3598: void ndv_print(NDV p)
3599: {
3600: NMV m;
3601: int i,len;
3602:
1.34 noro 3603: if ( !p ) printf("0\n");
1.3 noro 3604: else {
1.14 noro 3605: len = LEN(p);
1.3 noro 3606: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3607: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3608: else printf("+%d*",CM(m));
1.16 noro 3609: ndl_print(DL(m));
3610: }
3611: printf("\n");
3612: }
3613: }
3614:
3615: void ndv_print_q(NDV p)
3616: {
3617: NMV m;
3618: int i,len;
3619:
1.34 noro 3620: if ( !p ) printf("0\n");
1.16 noro 3621: else {
3622: len = LEN(p);
3623: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3624: printf("+");
1.74 noro 3625: printexpr(CO,(Obj)CQ(m));
1.16 noro 3626: printf("*");
1.14 noro 3627: ndl_print(DL(m));
1.3 noro 3628: }
3629: printf("\n");
3630: }
1.25 noro 3631: }
3632:
1.61 noro 3633: NODE ndv_reducebase(NODE x)
1.27 noro 3634: {
3635: int len,i,j;
3636: NDV *w;
3637: NODE t,t0;
3638:
3639: len = length(x);
3640: w = (NDV *)ALLOCA(len*sizeof(NDV));
3641: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3642: for ( i = 0; i < len; i++ ) {
3643: for ( j = 0; j < i; j++ ) {
3644: if ( w[i] && w[j] )
3645: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3646: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3647: }
3648: }
3649: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3650: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3651: }
3652: NEXT(t) = 0; x = t0;
3653: return x;
1.11 noro 3654: }
1.32 noro 3655:
1.43 noro 3656: /* XXX incomplete */
3657:
1.32 noro 3658: void nd_init_ord(struct order_spec *ord)
3659: {
1.43 noro 3660: switch ( ord->id ) {
1.32 noro 3661: case 0:
1.43 noro 3662: switch ( ord->ord.simple ) {
3663: case 0:
3664: nd_dcomp = 1;
3665: nd_isrlex = 1;
3666: break;
3667: case 1:
3668: nd_dcomp = 1;
3669: nd_isrlex = 0;
3670: break;
3671: case 2:
3672: nd_dcomp = 0;
3673: nd_isrlex = 0;
1.45 noro 3674: ndl_compare_function = ndl_lex_compare;
1.58 noro 3675: break;
3676: case 11:
3677: /* XXX */
3678: nd_dcomp = 0;
3679: nd_isrlex = 1;
3680: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3681: break;
3682: default:
3683: error("nd_gr : unsupported order");
3684: }
1.32 noro 3685: break;
3686: case 1:
1.43 noro 3687: /* XXX */
3688: nd_dcomp = -1;
1.32 noro 3689: nd_isrlex = 0;
1.45 noro 3690: ndl_compare_function = ndl_block_compare;
1.34 noro 3691: break;
1.43 noro 3692: case 2:
3693: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3694: break;
3695: }
1.41 noro 3696: nd_ord = ord;
1.32 noro 3697: }
3698:
1.43 noro 3699: BlockMask nd_create_blockmask(struct order_spec *ord)
3700: {
3701: int n,i,j,s,l;
1.61 noro 3702: UINT *t;
1.43 noro 3703: BlockMask bm;
3704:
3705: if ( !ord->id )
3706: return 0;
3707: n = ord->ord.block.length;
3708: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3709: bm->n = n;
3710: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3711: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3712: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3713: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3714: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3715: l = bm->order_pair[i].length;
3716: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3717: }
3718: return bm;
1.57 noro 3719: }
3720:
3721: EPOS nd_create_epos(struct order_spec *ord)
3722: {
3723: int i,j,l,s,ord_l,ord_o;
3724: EPOS epos;
3725: struct order_pair *op;
3726:
3727: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3728: switch ( ord->id ) {
3729: case 0:
3730: if ( nd_isrlex ) {
3731: for ( i = 0; i < nd_nvar; i++ ) {
3732: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3733: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3734: }
3735: } else {
3736: for ( i = 0; i < nd_nvar; i++ ) {
3737: epos[i].i = nd_exporigin + i/nd_epw;
3738: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3739: }
3740: }
3741: break;
3742: case 1:
3743: /* block order */
3744: l = ord->ord.block.length;
3745: op = ord->ord.block.order_pair;
3746: for ( j = 0, s = 0; j < l; j++ ) {
3747: ord_o = op[j].order;
3748: ord_l = op[j].length;
3749: if ( !ord_o )
3750: for ( i = 0; i < ord_l; i++ ) {
3751: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3752: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3753: }
3754: else
3755: for ( i = 0; i < ord_l; i++ ) {
3756: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3757: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3758: }
3759: s += ord_l;
3760: }
3761: break;
3762: case 2:
3763: error("nd_create_epos : matrix order is not supported yet.");
3764: }
3765: return epos;
1.43 noro 3766: }
1.59 noro 3767:
3768: /* external interface */
3769:
3770: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3771: {
1.61 noro 3772: NODE t,in0,in;
1.59 noro 3773: ND nd,nf;
1.61 noro 3774: NDV ndv;
3775: VL vv,tv;
3776: int stat,nvar,max,e;
1.69 noro 3777: union oNDC dn;
1.59 noro 3778:
3779: pltovl(v,&vv);
1.74 noro 3780: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3781:
3782: /* get the degree bound */
3783: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3784: for ( tv = vv; tv; tv = NEXT(tv) ) {
3785: e = getdeg(tv->v,(P)BDY(t));
3786: max = MAX(e,max);
3787: }
3788: for ( tv = vv; tv; tv = NEXT(tv) ) {
3789: e = getdeg(tv->v,f);
3790: max = MAX(e,max);
3791: }
3792:
1.59 noro 3793: nd_init_ord(ord);
1.61 noro 3794: nd_setup_parameters(nvar,max);
3795:
3796: /* conversion to ndv */
3797: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3798: NEXTNODE(in0,in);
3799: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3800: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3801: }
3802: NEXTNODE(in0,in);
3803: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3804: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3805: NEXT(in) = 0;
3806:
3807: ndv_setup(m,0,in0);
1.59 noro 3808: nd_psn--;
3809: nd_scale=2;
3810: while ( 1 ) {
3811: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3812: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3813: if ( !stat ) {
3814: nd_psn++;
3815: nd_reconstruct(m,0,0);
3816: nd_psn--;
3817: } else
3818: break;
3819: }
1.61 noro 3820: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3821: }
3822:
3823: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3824: {
3825: NM m;
3826: UINT *t,*s;
3827: int i;
3828:
3829: for ( i = 0; i < n; i++ ) r[i] = 0;
3830: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3831: t = DL(m);
3832: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3833: r[i] = CM(m);
3834: }
3835: for ( i = 0; !r[i]; i++ );
3836: return i;
3837: }
3838:
1.74 noro 3839: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3840: {
3841: NMV m;
3842: UINT *t,*s;
3843: int i,j,len;
3844:
3845: for ( i = 0; i < n; i++ ) r[i] = 0;
3846: m = BDY(d);
3847: len = LEN(d);
3848: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3849: t = DL(m);
3850: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3851: r[i] = CM(m);
3852: }
3853: for ( i = 0; !r[i]; i++ );
3854: return i;
3855: }
3856:
1.63 noro 3857: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3858: {
3859: NM m;
3860: NMV mr;
3861: UINT *d,*t,*s;
3862: NDV p;
3863: int i,j,len;
3864:
3865: m = pair->mul;
3866: d = DL(m);
3867: p = nd_ps[pair->index];
3868: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3869: for ( i = 0; i < n; i++ ) r[i] = 0;
3870: len = LEN(p);
3871: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3872: ndl_add(d,DL(mr),t);
3873: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3874: r[i] = CM(mr);
3875: }
3876: for ( i = 0; !r[i]; i++ );
3877: return i;
3878: }
3879:
1.67 noro 3880: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3881: {
3882: NM m;
3883: NMV mr;
3884: UINT *d,*t,*s;
3885: NDV p;
1.67 noro 3886: unsigned char *ivc;
3887: unsigned short *ivs;
1.81 noro 3888: UINT *v,*ivi,*s0v;
1.67 noro 3889: int i,j,len,prev,diff,cdiff;
3890: IndArray r;
1.64 noro 3891:
3892: m = pair->mul;
3893: d = DL(m);
3894: p = nd_ps[pair->index];
3895: len = LEN(p);
3896: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3897: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3898: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3899: ndl_add(d,DL(mr),t);
3900: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3901: v[j] = i;
3902: }
1.81 noro 3903: r = (IndArray)MALLOC(sizeof(struct oIndArray));
1.67 noro 3904: r->head = v[0];
3905: diff = 0;
3906: for ( i = 1; i < len; i++ ) {
3907: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3908: }
3909: if ( diff < 256 ) {
3910: r->width = 1;
3911: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3912: r->index.c = ivc;
3913: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3914: } else if ( diff < 65536 ) {
3915: r->width = 2;
3916: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3917: r->index.s = ivs;
3918: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3919: } else {
3920: r->width = 4;
3921: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3922: r->index.i = ivi;
3923: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3924: }
1.67 noro 3925: return r;
1.64 noro 3926: }
3927:
1.65 noro 3928:
1.76 noro 3929: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3930: {
1.67 noro 3931: int i,j,k,len,pos,prev;
1.66 noro 3932: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3933: IndArray ivect;
3934: unsigned char *ivc;
3935: unsigned short *ivs;
3936: unsigned int *ivi;
1.65 noro 3937: NDV redv;
1.67 noro 3938: NMV mr;
1.65 noro 3939: NODE rp;
1.76 noro 3940: int maxrs;
1.65 noro 3941:
1.76 noro 3942: maxrs = 0;
1.74 noro 3943: for ( i = 0; i < nred; i++ ) {
1.65 noro 3944: ivect = imat[i];
1.67 noro 3945: k = ivect->head; svect[k] %= m;
1.65 noro 3946: if ( c = svect[k] ) {
1.76 noro 3947: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3948: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3949: len = LEN(redv); mr = BDY(redv);
3950: svect[k] = 0; prev = k;
3951: switch ( ivect->width ) {
3952: case 1:
3953: ivc = ivect->index.c;
3954: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3955: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3956: prev = pos;
3957: DMA(c1,c,c2,up,lo);
3958: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3959: } else svect[pos] = lo;
3960: }
3961: break;
3962: case 2:
3963: ivs = ivect->index.s;
3964: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3965: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3966: prev = pos;
3967: DMA(c1,c,c2,up,lo);
3968: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3969: } else svect[pos] = lo;
3970: }
3971: break;
3972: case 4:
3973: ivi = ivect->index.i;
3974: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3975: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3976: prev = pos;
3977: DMA(c1,c,c2,up,lo);
3978: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3979: } else svect[pos] = lo;
3980: }
3981: break;
1.65 noro 3982: }
3983: }
3984: }
1.66 noro 3985: for ( i = 0; i < col; i++ )
3986: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 noro 3987: return maxrs;
1.65 noro 3988: }
3989:
1.76 noro 3990: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3991: {
3992: int i,j,k,len,pos,prev;
3993: UINT c,c1,c2,c3,up,lo,dmy;
3994: IndArray ivect;
3995: unsigned char *ivc;
3996: unsigned short *ivs;
3997: unsigned int *ivi;
3998: NDV redv;
3999: NMV mr;
4000: NODE rp;
1.76 noro 4001: int maxrs;
1.72 noro 4002:
1.76 noro 4003: maxrs = 0;
1.74 noro 4004: for ( i = 0; i < nred; i++ ) {
1.72 noro 4005: ivect = imat[i];
4006: k = ivect->head; svect[k] %= m;
4007: if ( c = svect[k] ) {
1.76 noro 4008: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 4009: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 4010: len = LEN(redv); mr = BDY(redv);
4011: svect[k] = 0; prev = k;
4012: switch ( ivect->width ) {
4013: case 1:
4014: ivc = ivect->index.c;
4015: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4016: pos = prev+ivc[j]; prev = pos;
4017: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4018: }
4019: break;
4020: case 2:
4021: ivs = ivect->index.s;
4022: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4023: pos = prev+ivs[j]; prev = pos;
4024: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4025: }
4026: break;
4027: case 4:
4028: ivi = ivect->index.i;
4029: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
4030: pos = prev+ivi[j]; prev = pos;
4031: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
4032: }
4033: break;
4034: }
4035: }
4036: }
1.76 noro 4037: return maxrs;
1.72 noro 4038: }
4039:
1.65 noro 4040: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
4041: {
4042: int j,k,len;
4043: UINT *p;
4044: UINT c;
4045: NDV r;
4046: NMV mr0,mr;
4047:
4048: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
4049: if ( !len ) return 0;
4050: else {
1.74 noro 4051: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4052: ndv_alloc += nmv_adv*len;
1.65 noro 4053: mr = mr0;
4054: p = s0vect;
4055: for ( j = k = 0; j < col; j++, p += nd_wpd )
4056: if ( !rhead[j] ) {
4057: if ( c = vect[k++] ) {
4058: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
4059: }
4060: }
4061: MKNDV(nd_nvar,mr0,len,r);
4062: return r;
4063: }
4064: }
4065:
1.74 noro 4066: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 4067: {
4068: ND_pairs t;
4069: NODE sp0,sp;
4070: int stat;
4071: ND spol;
4072:
4073: for ( t = l; t; t = NEXT(t) ) {
4074: stat = nd_sp(m,0,t,&spol);
4075: if ( !stat ) return 0;
4076: if ( spol ) {
4077: add_pbucket_symbolic(bucket,spol);
4078: }
4079: }
1.68 noro 4080: return 1;
1.65 noro 4081: }
4082:
4083: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
4084: {
4085: NODE rp0,rp;
4086: NM mul,head,s0,s;
1.76 noro 4087: int index,col,i,sugar;
1.65 noro 4088: RHist h;
4089: UINT *s0v,*p;
4090: NM_ind_pair pair;
4091: ND red;
4092:
4093: s0 = 0; rp0 = 0; col = 0;
4094: while ( 1 ) {
4095: head = remove_head_pbucket_symbolic(bucket);
4096: if ( !head ) break;
4097: if ( !s0 ) s0 = head;
4098: else NEXT(s) = head;
4099: s = head;
4100: index = ndl_find_reducer(DL(head));
4101: if ( index >= 0 ) {
4102: h = nd_psh[index];
4103: NEWNM(mul);
4104: ndl_sub(DL(head),DL(h),DL(mul));
4105: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 noro 4106: sugar = TD(DL(mul))+SG(nd_ps[index]);
4107: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 4108: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
4109: add_pbucket_symbolic(bucket,nd_remove_head(red));
4110: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
4111: }
4112: col++;
4113: }
1.72 noro 4114: if ( rp0 ) NEXT(rp) = 0;
4115: NEXT(s) = 0;
1.65 noro 4116: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
4117: for ( i = 0, p = s0v, s = s0; i < col;
4118: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
4119: *s0vect = s0v;
4120: *r = rp0;
4121: return col;
4122: }
4123:
1.69 noro 4124: NODE nd_f4(int m)
4125: {
4126: int i,nh,stat,index;
4127: NODE r,g;
4128: ND_pairs d,l,t;
4129: ND spol,red;
4130: NDV nf,redv;
4131: NM s0,s;
1.74 noro 4132: NODE rp0,sp0,srp0,nflist;
1.69 noro 4133: int nsp,nred,col,rank,len,k,j,a;
4134: UINT c;
1.74 noro 4135: UINT **spmat;
1.69 noro 4136: UINT *s0vect,*svect,*p,*v;
4137: int *colstat;
4138: IndArray *imat;
4139: int *rhead;
4140: int spcol,sprow;
4141: int sugar;
4142: PGeoBucket bucket;
4143: struct oEGT eg0,eg1,eg_f4;
4144:
4145: if ( !m )
4146: error("nd_f4 : not implemented");
1.74 noro 4147: ndv_alloc = 0;
1.69 noro 4148: g = 0; d = 0;
4149: for ( i = 0; i < nd_psn; i++ ) {
4150: d = update_pairs(d,g,i);
4151: g = update_base(g,i);
4152: }
4153: while ( d ) {
4154: get_eg(&eg0);
4155: l = nd_minsugarp(d,&d);
4156: sugar = SG(l);
4157: bucket = create_pbucket();
1.74 noro 4158: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4159: if ( !stat ) {
4160: for ( t = l; NEXT(t); t = NEXT(t) );
4161: NEXT(t) = d; d = l;
4162: d = nd_reconstruct(m,0,d);
4163: continue;
4164: }
4165: if ( !sp0 ) continue;
4166: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4167: if ( !col ) {
4168: for ( t = l; NEXT(t); t = NEXT(t) );
4169: NEXT(t) = d; d = l;
4170: d = nd_reconstruct(m,0,d);
4171: continue;
4172: }
4173: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4174: if ( DP_Print )
4175: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4176: sugar,eg_f4.exectime+eg_f4.gctime);
4177: if ( 1 )
4178: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4179: else
4180: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4181: /* adding new bases */
1.74 noro 4182: for ( r = nflist; r; r = NEXT(r) ) {
4183: nf = (NDV)BDY(r);
1.69 noro 4184: ndv_removecont(m,nf);
1.77 noro 4185: nh = ndv_newps(m,nf,0);
1.69 noro 4186: d = update_pairs(d,g,nh);
4187: g = update_base(g,nh);
4188: }
4189: }
4190: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 4191: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 4192: return g;
4193: }
1.74 noro 4194:
4195: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4196: {
1.67 noro 4197: IndArray *imat;
1.74 noro 4198: int nsp,nred,spcol,sprow,a;
1.65 noro 4199: int *rhead;
1.74 noro 4200: int i,j,k,l,rank;
4201: NODE rp,r0,r;
4202: ND_pairs sp;
4203: ND spol;
4204: int **spmat;
4205: UINT *svect,*v;
4206: int *colstat;
1.80 noro 4207: struct oEGT eg0,eg1,eg2,eg_f4,eg_f4_1,eg_f4_2;
1.74 noro 4208: NM_ind_pair *rvect;
1.76 noro 4209: int maxrs;
4210: int *spsugar;
1.63 noro 4211:
1.74 noro 4212: get_eg(&eg0);
4213: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4214: nred = length(rp0); spcol = col-nred;
4215: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4216: rhead = (int *)ALLOCA(col*sizeof(int));
4217: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4218:
1.74 noro 4219: /* construction of index arrays */
4220: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4221: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4222: rvect[i] = (NM_ind_pair)BDY(rp);
4223: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4224: rhead[imat[i]->head] = 1;
4225: }
4226:
4227: /* elimination (1st step) */
4228: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4229: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 noro 4230: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4231: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4232: nd_sp(m,0,sp,&spol);
1.75 noro 4233: if ( !spol ) continue;
1.74 noro 4234: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4235: if ( m == -1 )
4236: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
4237: else
4238: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4239: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4240: if ( i < col ) {
4241: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4242: for ( j = k = 0; j < col; j++ )
4243: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4244: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4245: sprow++;
4246: }
1.76 noro 4247: nd_free(spol);
1.74 noro 4248: }
1.80 noro 4249: get_eg(&eg1); init_eg(&eg_f4_1); add_eg(&eg_f4_1,&eg0,&eg1);
4250: if ( DP_Print ) {
4251: fprintf(asir_out,"elim1=%fsec,",eg_f4_1.exectime+eg_f4_1.gctime);
4252: fflush(asir_out);
4253: }
1.74 noro 4254: /* free index arrays */
4255: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4256:
4257: /* elimination (2nd step) */
4258: colstat = (int *)ALLOCA(spcol*sizeof(int));
4259: if ( m == -1 )
1.76 noro 4260: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4261: else
1.76 noro 4262: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4263: r0 = 0;
4264: for ( i = 0; i < rank; i++ ) {
4265: NEXTNODE(r0,r); BDY(r) =
4266: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 noro 4267: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4268: GC_free(spmat[i]);
4269: }
4270: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
1.80 noro 4271: get_eg(&eg2); init_eg(&eg_f4_2); add_eg(&eg_f4_2,&eg1,&eg2);
4272: init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg2);
1.74 noro 4273: if ( DP_Print ) {
1.80 noro 4274: fprintf(asir_out,"elim2=%fsec\n",eg_f4_2.exectime+eg_f4_2.gctime);
1.74 noro 4275: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4276: nsp,nred,sprow,spcol,rank);
4277: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4278: }
1.74 noro 4279: return r0;
4280: }
4281:
4282: FILE *nd_write,*nd_read;
4283:
4284: void nd_send_int(int a) {
4285: write_int(nd_write,&a);
4286: }
4287:
4288: void nd_send_intarray(int *p,int len) {
4289: write_intarray(nd_write,p,len);
4290: }
4291:
4292: int nd_recv_int() {
4293: int a;
4294:
4295: read_int(nd_read,&a);
4296: return a;
4297: }
4298:
4299: void nd_recv_intarray(int *p,int len) {
4300: read_intarray(nd_read,p,len);
4301: }
4302:
4303: void nd_send_ndv(NDV p) {
4304: int len,i;
4305: NMV m;
4306:
4307: if ( !p ) nd_send_int(0);
4308: else {
4309: len = LEN(p);
4310: nd_send_int(len);
4311: m = BDY(p);
4312: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4313: nd_send_int(CM(m));
4314: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4315: }
1.74 noro 4316: }
4317: }
4318:
4319: void nd_send_nd(ND p) {
4320: int len,i;
4321: NM m;
4322:
4323: if ( !p ) nd_send_int(0);
4324: else {
4325: len = LEN(p);
4326: nd_send_int(len);
4327: m = BDY(p);
4328: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4329: nd_send_int(CM(m));
4330: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4331: }
1.74 noro 4332: }
4333: }
1.65 noro 4334:
1.74 noro 4335: NDV nd_recv_ndv()
4336: {
4337: int len,i;
4338: NMV m,m0;
4339: NDV r;
1.65 noro 4340:
1.74 noro 4341: len = nd_recv_int();
4342: if ( !len ) return 0;
4343: else {
4344: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4345: ndv_alloc += len*nmv_adv;
4346: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4347: CM(m) = nd_recv_int();
4348: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4349: }
1.74 noro 4350: MKNDV(nd_nvar,m0,len,r);
4351: return r;
4352: }
4353: }
1.65 noro 4354:
1.74 noro 4355: int ox_exec_f4_red(Q proc)
4356: {
4357: Obj obj;
4358: STRING fname;
4359: NODE arg;
4360: int s;
4361: extern int ox_need_conv,ox_file_io;
4362:
4363: MKSTR(fname,"nd_exec_f4_red");
4364: arg = mknode(2,proc,fname);
4365: Pox_cmo_rpc(arg,&obj);
4366: s = get_ox_server_id(QTOS(proc));
4367: nd_write = iofp[s].out;
4368: nd_read = iofp[s].in;
4369: ox_need_conv = ox_file_io = 0;
4370: return s;
4371: }
4372:
4373: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4374: {
4375: int nsp,nred;
4376: int i,rank,s;
4377: NODE rp,r0,r;
4378: ND_pairs sp;
4379: NM_ind_pair pair;
4380: NMV nmv;
4381: NM nm;
4382: NDV nf;
4383: Obj proc,dmy;
4384:
4385: ox_launch_main(0,0,&proc);
4386: s = ox_exec_f4_red((Q)proc);
4387:
4388: nd_send_int(m);
4389: nd_send_int(nd_nvar);
4390: nd_send_int(nd_bpe);
4391: nd_send_int(nd_wpd);
4392: nd_send_int(nmv_adv);
4393:
4394: saveobj(nd_write,dp_current_spec.obj); fflush(nd_write);
4395:
4396: nd_send_int(nd_psn);
4397: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4398:
4399: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4400: nd_send_int(nsp);
4401: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4402: nd_send_int(sp->i1); nd_send_int(sp->i2);
4403: }
4404:
4405: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4406:
4407: nred = length(rp0); nd_send_int(nred);
4408: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4409: pair = (NM_ind_pair)BDY(rp);
4410: nd_send_int(pair->index);
4411: nd_send_intarray(pair->mul->dl,nd_wpd);
4412: }
4413: fflush(nd_write);
4414: rank = nd_recv_int();
4415: fprintf(asir_out,"rank=%d\n",rank);
4416: r0 = 0;
4417: for ( i = 0; i < rank; i++ ) {
4418: nf = nd_recv_ndv();
4419: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4420: }
4421: Pox_shutdown(mknode(1,proc),&dmy);
4422: return r0;
4423: }
4424:
4425: /* server side */
4426:
4427: void nd_exec_f4_red_dist()
4428: {
4429: int m,i,nsp,col,s0size,nred,spcol,j,k;
4430: NM_ind_pair *rp0;
4431: NDV nf;
4432: UINT *s0vect;
4433: IndArray *imat;
4434: int *rhead;
4435: int **spmat;
4436: UINT *svect,*v;
4437: ND_pairs *sp0;
4438: int *colstat;
4439: int a,sprow,rank;
4440: struct order_spec ord;
4441: Obj ordspec;
4442: ND spol;
1.76 noro 4443: int maxrs;
4444: int *spsugar;
1.74 noro 4445:
4446: nd_read = iofp[0].in;
4447: nd_write = iofp[0].out;
4448: m = nd_recv_int();
4449: nd_nvar = nd_recv_int();
4450: nd_bpe = nd_recv_int();
4451: nd_wpd = nd_recv_int();
4452: nmv_adv = nd_recv_int();
4453:
4454: loadobj(nd_read,&ordspec);
4455: create_order_spec(ordspec,&ord);
4456: nd_init_ord(&ord);
4457: nd_setup_parameters(nd_nvar,0);
4458:
4459: nd_psn = nd_recv_int();
4460: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4461: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4462: for ( i = 0; i < nd_psn; i++ ) {
4463: nd_ps[i] = nd_recv_ndv();
4464: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4465: }
4466:
4467: nsp = nd_recv_int();
4468: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4469: for ( i = 0; i < nsp; i++ ) {
4470: NEWND_pairs(sp0[i]);
4471: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4472: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4473: }
4474:
4475: col = nd_recv_int();
4476: s0size = col*nd_wpd;
4477: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4478: nd_recv_intarray(s0vect,s0size);
4479:
4480: nred = nd_recv_int();
4481: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4482: for ( i = 0; i < nred; i++ ) {
4483: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4484: rp0[i]->index = nd_recv_int();
4485: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4486: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4487: }
4488:
4489: spcol = col-nred;
4490: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4491: rhead = (int *)MALLOC(col*sizeof(int));
4492: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4493:
4494: /* construction of index arrays */
4495: for ( i = 0; i < nred; i++ ) {
4496: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4497: rhead[imat[i]->head] = 1;
4498: }
4499:
4500: /* elimination (1st step) */
4501: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4502: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 noro 4503: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4504: for ( a = sprow = 0; a < nsp; a++ ) {
4505: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4506: if ( !spol ) continue;
1.74 noro 4507: nd_to_vect(m,s0vect,col,spol,svect);
1.76 noro 4508: if ( m == -1 )
4509: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
4510: else
4511: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4512: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4513: if ( i < col ) {
4514: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4515: for ( j = k = 0; j < col; j++ )
4516: if ( !rhead[j] ) v[k++] = svect[j];
1.76 noro 4517: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4518: sprow++;
4519: }
1.76 noro 4520: nd_free(spol);
1.74 noro 4521: }
4522: /* elimination (2nd step) */
4523: colstat = (int *)ALLOCA(spcol*sizeof(int));
4524: if ( m == -1 )
1.76 noro 4525: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4526: else
1.76 noro 4527: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4528: nd_send_int(rank);
4529: for ( i = 0; i < rank; i++ ) {
4530: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4531: nd_send_ndv(nf);
1.63 noro 4532: }
1.74 noro 4533: fflush(nd_write);
1.76 noro 4534: }
4535:
4536: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4537: {
4538: int i,j,k,l,inv,a,rank,s;
4539: unsigned int *t,*pivot,*pk;
4540: unsigned int **mat;
4541:
4542: mat = (unsigned int **)mat0;
4543: for ( rank = 0, j = 0; j < col; j++ ) {
4544: for ( i = rank; i < row; i++ )
4545: mat[i][j] %= md;
4546: for ( i = rank; i < row; i++ )
4547: if ( mat[i][j] )
4548: break;
4549: if ( i == row ) {
4550: colstat[j] = 0;
4551: continue;
4552: } else
4553: colstat[j] = 1;
4554: if ( i != rank ) {
4555: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4556: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4557: }
4558: pivot = mat[rank];
4559: s = sugar[rank];
4560: inv = invm(pivot[j],md);
4561: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4562: if ( *pk ) {
4563: if ( *pk >= (unsigned int)md )
4564: *pk %= md;
4565: DMAR(*pk,inv,0,md,*pk)
4566: }
4567: for ( i = rank+1; i < row; i++ ) {
4568: t = mat[i];
4569: if ( a = t[j] ) {
4570: sugar[i] = MAX(sugar[i],s);
4571: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4572: }
4573: }
4574: rank++;
4575: }
4576: for ( j = col-1, l = rank-1; j >= 0; j-- )
4577: if ( colstat[j] ) {
4578: pivot = mat[l];
4579: s = sugar[l];
4580: for ( i = 0; i < l; i++ ) {
4581: t = mat[i];
4582: t[j] %= md;
4583: if ( a = t[j] ) {
4584: sugar[i] = MAX(sugar[i],s);
4585: red_by_vect(md,t+j,pivot+j,md-a,col-j);
4586: }
4587: }
4588: l--;
4589: }
4590: for ( j = 0, l = 0; l < rank; j++ )
4591: if ( colstat[j] ) {
4592: t = mat[l];
4593: for ( k = j; k < col; k++ )
4594: if ( t[k] >= (unsigned int)md )
4595: t[k] %= md;
4596: l++;
4597: }
4598: return rank;
4599: }
4600:
4601: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
4602: {
4603: int i,j,k,l,inv,a,rank,s;
4604: unsigned int *t,*pivot,*pk;
4605: unsigned int **mat;
4606:
4607: mat = (unsigned int **)mat0;
4608: for ( rank = 0, j = 0; j < col; j++ ) {
4609: for ( i = rank; i < row; i++ )
4610: if ( mat[i][j] )
4611: break;
4612: if ( i == row ) {
4613: colstat[j] = 0;
4614: continue;
4615: } else
4616: colstat[j] = 1;
4617: if ( i != rank ) {
4618: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
4619: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
4620: }
4621: pivot = mat[rank];
4622: s = sugar[rank];
4623: inv = _invsf(pivot[j]);
4624: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
4625: if ( *pk )
4626: *pk = _mulsf(*pk,inv);
4627: for ( i = rank+1; i < row; i++ ) {
4628: t = mat[i];
4629: if ( a = t[j] ) {
4630: sugar[i] = MAX(sugar[i],s);
4631: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4632: }
4633: }
4634: rank++;
4635: }
4636: for ( j = col-1, l = rank-1; j >= 0; j-- )
4637: if ( colstat[j] ) {
4638: pivot = mat[l];
4639: s = sugar[l];
4640: for ( i = 0; i < l; i++ ) {
4641: t = mat[i];
4642: if ( a = t[j] ) {
4643: sugar[i] = MAX(sugar[i],s);
4644: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
4645: }
4646: }
4647: l--;
4648: }
4649: return rank;
1.77 noro 4650: }
4651:
4652: int ndv_ishomo(NDV p)
4653: {
4654: NMV m;
4655: int len,h;
4656:
4657: if ( !p ) return 1;
4658: len = LEN(p);
4659: m = BDY(p);
4660: h = TD(DL(m));
4661: NMV_ADV(m);
4662: for ( len--; len; len--, NMV_ADV(m) )
4663: if ( TD(DL(m)) != h ) return 0;
4664: return 1;
4665: }
4666:
4667: void ndv_save(NDV p,int index)
4668: {
4669: FILE *s;
4670: char name[BUFSIZ];
4671: short id;
4672: int nv,sugar,len,n,i,td,e,j;
4673: NMV m;
4674: unsigned int *dl;
4675:
4676: sprintf(name,"%s/%d",Demand,index);
4677: s = fopen(name,"w");
4678: savevl(s,0);
4679: if ( !p ) {
4680: saveobj(s,0);
4681: return;
4682: }
4683: id = O_DP;
4684: nv = NV(p);
4685: sugar = SG(p);
4686: len = LEN(p);
4687: write_short(s,&id); write_int(s,&nv); write_int(s,&sugar);
4688: write_int(s,&len);
4689:
4690: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
4691: saveobj(s,(Obj)CQ(m));
4692: dl = DL(m);
4693: td = TD(dl);
4694: write_int(s,&td);
4695: for ( j = 0; j < nv; j++ ) {
4696: e = GET_EXP(dl,j);
4697: write_int(s,&e);
4698: }
4699: }
4700: fclose(s);
4701: }
4702:
4703: NDV ndv_load(int index)
4704: {
4705: FILE *s;
4706: char name[BUFSIZ];
4707: short id;
4708: int nv,sugar,len,n,i,td,e,j;
4709: NDV d;
4710: NMV m0,m;
4711: unsigned int *dl;
4712: Obj obj;
4713:
4714: sprintf(name,"%s/%d",Demand,index);
4715: s = fopen(name,"r");
4716: if ( !s ) return 0;
4717:
4718: skipvl(s);
4719: read_short(s,&id);
4720: if ( !id ) return 0;
4721: read_int(s,&nv);
4722: read_int(s,&sugar);
4723: read_int(s,&len);
4724:
4725: m0 = m = MALLOC(len*nmv_adv);
4726: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4727: loadobj(s,&obj); CQ(m) = (Q)obj;
4728: dl = DL(m);
4729: ndl_zero(dl);
4730: read_int(s,&td); TD(dl) = td;
4731: for ( j = 0; j < nv; j++ ) {
4732: read_int(s,&e);
4733: PUT_EXP(dl,j,e);
4734: }
4735: if ( nd_blockmask ) ndl_weight_mask(dl);
4736: }
4737: fclose(s);
4738: MKNDV(nv,m0,len,d);
4739: SG(d) = sugar;
4740: return d;
1.59 noro 4741: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>