Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.76
1.76 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.75 2003/09/19 10:09:42 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;
1.1 noro 147:
1.74 noro 148: extern struct order_spec dp_current_spec;
1.61 noro 149: extern VL CO;
1.71 noro 150: extern int Top,Reverse,DP_Print,dp_nelim,do_weyl;
1.58 noro 151: extern int *current_weyl_weight_vector;
1.1 noro 152:
1.40 noro 153: /* fundamental macros */
1.34 noro 154: #define TD(d) (d[0])
1.1 noro 155: #define HDL(d) ((d)->body->dl)
1.34 noro 156: #define HTD(d) (TD(HDL(d)))
1.14 noro 157: #define HCM(d) ((d)->body->c.m)
1.16 noro 158: #define HCQ(d) ((d)->body->c.z)
1.61 noro 159: #define HCP(d) ((d)->body->c.p)
1.14 noro 160: #define CM(a) ((a)->c.m)
1.16 noro 161: #define CQ(a) ((a)->c.z)
1.61 noro 162: #define CP(a) ((a)->c.p)
1.14 noro 163: #define DL(a) ((a)->dl)
164: #define SG(a) ((a)->sugar)
165: #define LEN(a) ((a)->len)
1.33 noro 166: #define LCM(a) ((a)->lcm)
1.42 noro 167: #define GET_EXP(d,a) (((d)[nd_epos[a].i]>>nd_epos[a].s)&nd_mask0)
1.60 noro 168: #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 169: #define PUT_EXP(r,a,e) ((r)[nd_epos[a].i] |= ((e)<<nd_epos[a].s))
1.45 noro 170: #define XOR_EXP(r,a,e) ((r)[nd_epos[a].i] ^= ((e)<<nd_epos[a].s))
1.1 noro 171:
1.61 noro 172: #define GET_EXP_OLD(d,a) (((d)[oepos[a].i]>>oepos[a].s)&omask0)
173: #define PUT_EXP_OLD(r,a,e) ((r)[oepos[a].i] |= ((e)<<oepos[a].s))
174:
1.40 noro 175: /* macros for term comparison */
1.34 noro 176: #define TD_DL_COMPARE(d1,d2)\
1.41 noro 177: (TD(d1)>TD(d2)?1:(TD(d1)<TD(d2)?-1:ndl_lex_compare(d1,d2)))
1.43 noro 178: #if 0
1.34 noro 179: #define DL_COMPARE(d1,d2)\
1.43 noro 180: (nd_dcomp>0?TD_DL_COMPARE(d1,d2)\
181: :(nd_dcomp==0?ndl_lex_compare(d1,d2)\
182: :(nd_blockmask?ndl_block_compare(d1,d2)\
1.45 noro 183: :(*ndl_compare_function)(d1,d2))))
1.43 noro 184: #else
185: #define DL_COMPARE(d1,d2)\
1.45 noro 186: (nd_dcomp>0?TD_DL_COMPARE(d1,d2):(*ndl_compare_function)(d1,d2))
1.43 noro 187: #endif
1.34 noro 188:
1.40 noro 189: /* allocators */
1.15 noro 190: #define NEWRHist(r) \
1.61 noro 191: ((r)=(RHist)MALLOC(sizeof(struct oRHist)+(nd_wpd-1)*sizeof(UINT)))
1.34 noro 192: #define NEWND_pairs(m) \
193: if(!_ndp_free_list)_NDP_alloc();\
194: (m)=_ndp_free_list; _ndp_free_list = NEXT(_ndp_free_list)
195: #define NEWNM(m)\
196: if(!_nm_free_list)_NM_alloc();\
197: (m)=_nm_free_list; _nm_free_list = NEXT(_nm_free_list)
198: #define MKND(n,m,len,d)\
199: if(!_nd_free_list)_ND_alloc();\
200: (d)=_nd_free_list; _nd_free_list = (ND)BDY(_nd_free_list);\
201: NV(d)=(n); LEN(d)=(len); BDY(d)=(m)
1.40 noro 202: #define NEWNDV(d) ((d)=(NDV)MALLOC(sizeof(struct oNDV)))
203: #define MKNDV(n,m,l,d) NEWNDV(d); NV(d)=(n); BDY(d)=(m); LEN(d) = l;
1.63 noro 204: #define NEWNM_ind_pair(p)\
205: ((p)=(NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair)))
1.1 noro 206:
1.40 noro 207: /* allocate and link a new object */
1.13 noro 208: #define NEXTRHist(r,c) \
209: if(!(r)){NEWRHist(r);(c)=(r);}else{NEWRHist(NEXT(c));(c)=NEXT(c);}
1.1 noro 210: #define NEXTNM(r,c) \
211: if(!(r)){NEWNM(r);(c)=(r);}else{NEWNM(NEXT(c));(c)=NEXT(c);}
212: #define NEXTNM2(r,c,s) \
213: if(!(r)){(c)=(r)=(s);}else{NEXT(c)=(s);(c)=(s);}
1.40 noro 214: #define NEXTND_pairs(r,c) \
215: if(!(r)){NEWND_pairs(r);(c)=(r);}else{NEWND_pairs(NEXT(c));(c)=NEXT(c);}
1.76 ! noro 216: #define MKNM_ind_pair(p,m,i,s) (NEWNM_ind_pair(p),(p)->mul=(m),(p)->index=(i),(p)->sugar = (s))
1.34 noro 217:
1.40 noro 218: /* deallocators */
1.1 noro 219: #define FREENM(m) NEXT(m)=_nm_free_list; _nm_free_list=(m)
220: #define FREENDP(m) NEXT(m)=_ndp_free_list; _ndp_free_list=(m)
221: #define FREEND(m) BDY(m)=(NM)_nd_free_list; _nd_free_list=(m)
222:
1.40 noro 223: /* macro for increasing pointer to NMV */
224: #define NMV_ADV(m) (m = (NMV)(((char *)m)+nmv_adv))
1.61 noro 225: #define NMV_OADV(m) (m = (NMV)(((char *)m)+oadv))
226: #define NDV_NADV(m) (m = (NMV)(((char *)m)+newadv))
1.56 noro 227: #define NMV_PREV(m) (m = (NMV)(((char *)m)-nmv_adv))
1.61 noro 228: #define NMV_OPREV(m) (m = (NMV)(((char *)m)-oadv))
229:
1.40 noro 230: /* external functions */
1.71 noro 231: #if 1
1.40 noro 232: void GC_gcollect();
1.71 noro 233: #endif
1.40 noro 234: NODE append_one(NODE,int);
1.1 noro 235:
1.40 noro 236: /* manipulation of coefficients */
1.20 noro 237: void nd_removecont(int mod,ND p);
1.21 noro 238: void nd_removecont2(ND p1,ND p2);
1.40 noro 239: void removecont_array(Q *c,int n);
240:
241: /* GeoBucket functions */
1.25 noro 242: ND normalize_pbucket(int mod,PGeoBucket g);
243: int head_pbucket(int mod,PGeoBucket g);
1.26 noro 244: int head_pbucket_q(PGeoBucket g);
1.63 noro 245: void add_pbucket_symbolic(PGeoBucket g,ND d);
1.31 noro 246: void add_pbucket(int mod,PGeoBucket g,ND d);
1.25 noro 247: void free_pbucket(PGeoBucket b);
1.26 noro 248: void mulq_pbucket(PGeoBucket g,Q c);
1.63 noro 249: NM remove_head_pbucket_symbolic(PGeoBucket g);
1.25 noro 250: PGeoBucket create_pbucket();
1.20 noro 251:
1.40 noro 252: /* manipulation of pairs and bases */
1.39 noro 253: int nd_newps(int mod,ND a,ND aq);
1.40 noro 254: ND_pairs nd_newpairs( NODE g, int t );
1.1 noro 255: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest );
1.63 noro 256: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest );
1.1 noro 257: NODE update_base(NODE nd,int ndp);
1.40 noro 258: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t);
259: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest );
260: ND_pairs crit_B( ND_pairs d, int s );
261: ND_pairs crit_M( ND_pairs d1 );
262: ND_pairs crit_F( ND_pairs d1 );
1.1 noro 263: int crit_2( int dp1, int dp2 );
1.63 noro 264: int ndv_newps(NDV a,NDV aq);
1.40 noro 265:
266: /* top level functions */
1.63 noro 267: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp);
1.52 noro 268: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp);
1.63 noro 269: NODE nd_f4(int m);
1.27 noro 270: NODE nd_gb(int m,int checkonly);
1.23 noro 271: NODE nd_gb_trace(int m);
1.40 noro 272:
273: /* ndl functions */
1.61 noro 274: int ndl_weight(UINT *d);
275: void ndl_weight_mask(UINT *d);
276: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight);
277: void ndl_dehomogenize(UINT *p);
278: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos);
279: INLINE int ndl_reducible(UINT *d1,UINT *d2);
280: INLINE int ndl_lex_compare(UINT *d1,UINT *d2);
281: INLINE int ndl_block_compare(UINT *d1,UINT *d2);
282: INLINE int ndl_equal(UINT *d1,UINT *d2);
283: INLINE void ndl_copy(UINT *d1,UINT *d2);
284: INLINE void ndl_zero(UINT *d);
285: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d);
286: INLINE void ndl_addto(UINT *d1,UINT *d2);
287: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d);
288: INLINE int ndl_hash_value(UINT *d);
1.45 noro 289:
290: /* normal forms */
1.63 noro 291: INLINE int ndl_find_reducer(UINT *g);
1.53 noro 292: int nd_sp(int mod,int trace,ND_pairs p,ND *nf);
1.69 noro 293: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *nf);
1.53 noro 294: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *nf);
1.40 noro 295:
296: /* finalizers */
1.61 noro 297: NODE ndv_reducebase(NODE x);
298: NODE ndv_reduceall(int m,NODE f);
1.40 noro 299:
300: /* allocators */
301: void nd_free_private_storage();
302: void _NM_alloc();
303: void _ND_alloc();
1.1 noro 304: void nd_free(ND p);
1.40 noro 305: void nd_free_redlist();
306:
307: /* printing */
1.61 noro 308: void ndl_print(UINT *dl);
1.1 noro 309: void nd_print(ND p);
1.16 noro 310: void nd_print_q(ND p);
1.1 noro 311: void ndp_print(ND_pairs d);
1.40 noro 312:
313:
314: /* setup, reconstruct */
315: void nd_init_ord(struct order_spec *spec);
316: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs ndp);
1.61 noro 317: void ndv_setup(int mod,int trace,NODE f);
318: void nd_setup_parameters(int nvar,int max);
1.43 noro 319: BlockMask nd_create_blockmask(struct order_spec *ord);
1.57 noro 320: EPOS nd_create_epos(struct order_spec *ord);
1.48 noro 321: int nd_get_exporigin(struct order_spec *ord);
1.61 noro 322: void ndv_mod(int mod,NDV p);
323: NDV ndv_dup(int mod,NDV p);
1.63 noro 324: ND nd_dup(ND p);
1.40 noro 325:
326: /* ND functions */
1.61 noro 327: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand);
1.40 noro 328: void nd_mul_c(int mod,ND p,int mul);
329: void nd_mul_c_q(ND p,Q mul);
1.61 noro 330: void nd_mul_c_p(VL vl,ND p,P mul);
1.40 noro 331: ND nd_remove_head(ND p);
1.69 noro 332: ND nd_separate_head(ND p,ND *head);
1.1 noro 333: int nd_length(ND p);
1.61 noro 334: void nd_append_red(UINT *d,int i);
335: UINT *ndv_compute_bound(NDV p);
1.6 noro 336: ND nd_copy(ND p);
1.63 noro 337: ND nd_merge(ND p1,ND p2);
1.40 noro 338: ND nd_add(int mod,ND p1,ND p2);
339: ND nd_add_q(ND p1,ND p2);
1.71 noro 340: ND nd_add_sf(ND p1,ND p2);
1.41 noro 341: INLINE int nd_length(ND p);
1.74 noro 342: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
343: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0);
1.4 noro 344:
1.40 noro 345: /* NDV functions */
1.55 noro 346: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p);
347: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen);
1.19 noro 348: void ndv_mul_c(int mod,NDV p,int mul);
1.40 noro 349: void ndv_mul_c_q(NDV p,Q mul);
1.63 noro 350: ND ndv_mul_nm_symbolic(NM m0,NDV p);
1.61 noro 351: ND ndv_mul_nm(int mod,NM m0,NDV p);
1.43 noro 352: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos);
1.61 noro 353: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos);
354: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS eops);
1.45 noro 355: void ndv_dehomogenize(NDV p,struct order_spec *spec);
1.40 noro 356: void ndv_removecont(int mod,NDV p);
357: void ndv_print(NDV p);
358: void ndv_print_q(NDV p);
359: void ndv_free(NDV p);
360:
361: /* converters */
1.61 noro 362: ND ptond(VL vl,VL dvl,P p);
363: NDV ptondv(VL vl,VL dvl,P p);
364: P ndvtop(int mod,VL vl,VL dvl,NDV p);
1.16 noro 365: NDV ndtondv(int mod,ND p);
1.23 noro 366: ND ndvtond(int mod,NDV p);
1.63 noro 367: int nm_ind_pair_to_vect(int m,UINT *s0,int n,NM_ind_pair pair,UINT *r);
1.67 noro 368: IndArray nm_ind_pair_to_vect_compress(int m,UINT *s0,int n,NM_ind_pair pair);
1.63 noro 369: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r);
1.1 noro 370:
1.76 ! noro 371: /* elimination */
! 372: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat);
! 373: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat);
! 374:
1.1 noro 375: void nd_free_private_storage()
376: {
377: _nm_free_list = 0;
1.5 noro 378: _ndp_free_list = 0;
1.71 noro 379: #if 0
1.1 noro 380: GC_gcollect();
1.71 noro 381: #endif
1.1 noro 382: }
383:
384: void _NM_alloc()
385: {
386: NM p;
387: int i;
388:
1.11 noro 389: for ( i = 0; i < 1024; i++ ) {
1.61 noro 390: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 391: p->next = _nm_free_list; _nm_free_list = p;
392: }
393: }
394:
395: void _ND_alloc()
396: {
397: ND p;
398: int i;
399:
400: for ( i = 0; i < 1024; i++ ) {
401: p = (ND)GC_malloc(sizeof(struct oND));
402: p->body = (NM)_nd_free_list; _nd_free_list = p;
403: }
404: }
405:
406: void _NDP_alloc()
407: {
408: ND_pairs p;
409: int i;
410:
1.11 noro 411: for ( i = 0; i < 1024; i++ ) {
1.1 noro 412: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 413: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 414: p->next = _ndp_free_list; _ndp_free_list = p;
415: }
416: }
417:
1.30 noro 418: INLINE int nd_length(ND p)
1.1 noro 419: {
420: NM m;
421: int i;
422:
423: if ( !p )
424: return 0;
425: else {
426: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
427: return i;
428: }
429: }
430:
1.61 noro 431: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 432: {
1.61 noro 433: UINT u1,u2;
1.1 noro 434: int i,j;
435:
1.34 noro 436: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 437: #if USE_UNROLL
1.1 noro 438: switch ( nd_bpe ) {
1.62 noro 439: case 3:
440: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
441: u1 = d1[i]; u2 = d2[i];
442: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
443: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
444: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
445: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
446: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
447: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
448: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
449: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
450: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
451: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
452: }
453: return 1;
454: break;
1.1 noro 455: case 4:
1.41 noro 456: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 457: u1 = d1[i]; u2 = d2[i];
458: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 459: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
460: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
461: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
462: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
463: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
464: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
465: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 466: }
467: return 1;
468: break;
469: case 6:
1.41 noro 470: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 471: u1 = d1[i]; u2 = d2[i];
472: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 473: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
474: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
475: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
476: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 477: }
478: return 1;
479: break;
480: case 8:
1.41 noro 481: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 482: u1 = d1[i]; u2 = d2[i];
483: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 484: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
485: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
486: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 487: }
488: return 1;
489: break;
490: case 16:
1.41 noro 491: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 492: u1 = d1[i]; u2 = d2[i];
493: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 494: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 495: }
496: return 1;
497: break;
498: case 32:
1.41 noro 499: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 500: if ( d1[i] < d2[i] ) return 0;
501: return 1;
502: break;
503: default:
1.41 noro 504: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 505: u1 = d1[i]; u2 = d2[i];
506: for ( j = 0; j < nd_epw; j++ )
507: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
508: }
509: return 1;
510: }
1.65 noro 511: #else
512: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
513: u1 = d1[i]; u2 = d2[i];
514: for ( j = 0; j < nd_epw; j++ )
515: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
516: }
517: return 1;
518: #endif
1.1 noro 519: }
520:
1.61 noro 521: /*
522: * If the current order is a block order,
523: * then the last block is length 1 and contains
524: * the homo variable. Otherwise, the original
525: * order is either 0 or 2.
526: */
527:
528: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 529: {
1.61 noro 530: int w,i,e,n,omask0;
531:
532: omask0 = (1<<obpe)-1;
533: n = nd_nvar-1;
534: ndl_zero(r);
535: for ( i = 0; i < n; i++ ) {
536: e = GET_EXP_OLD(d,i);
537: PUT_EXP(r,i,e);
538: }
539: w = TD(d);
540: PUT_EXP(r,nd_nvar-1,weight-w);
541: TD(r) = weight;
542: if ( nd_blockmask ) ndl_weight_mask(r);
543: }
544:
545: void ndl_dehomogenize(UINT *d)
546: {
547: UINT mask;
548: UINT h;
1.31 noro 549: int i,bits;
1.23 noro 550:
1.44 noro 551: if ( nd_blockmask ) {
552: h = GET_EXP(d,nd_nvar-1);
1.45 noro 553: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 554: TD(d) -= h;
555: d[nd_exporigin-1] -= h;
556: } else {
557: if ( nd_isrlex ) {
558: if ( nd_bpe == 32 ) {
559: h = d[nd_exporigin];
560: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
561: d[i-1] = d[i];
562: d[i-1] = 0;
563: TD(d) -= h;
564: } else {
565: bits = nd_epw*nd_bpe;
566: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
567: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
568: for ( i = nd_exporigin; i < nd_wpd; i++ )
569: d[i] = ((d[i]<<nd_bpe)&mask)
570: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
571: TD(d) -= h;
572: }
1.45 noro 573: } else {
574: h = GET_EXP(d,nd_nvar-1);
575: XOR_EXP(d,nd_nvar-1,h);
576: TD(d) -= h;
577: }
1.44 noro 578: }
1.23 noro 579: }
580:
1.61 noro 581: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 582: {
1.61 noro 583: UINT t1,t2,u,u1,u2;
1.43 noro 584: int i,j,l;
1.1 noro 585:
1.65 noro 586: #if USE_UNROLL
1.1 noro 587: switch ( nd_bpe ) {
1.62 noro 588: case 3:
589: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
590: u1 = d1[i]; u2 = d2[i];
591: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
592: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
593: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
594: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
595: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
596: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
597: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
598: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
599: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
600: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
601: d[i] = u;
602: }
603: break;
1.1 noro 604: case 4:
1.41 noro 605: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 606: u1 = d1[i]; u2 = d2[i];
607: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 608: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
609: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
610: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
611: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
612: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
613: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
614: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 615: d[i] = u;
616: }
617: break;
618: case 6:
1.41 noro 619: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 620: u1 = d1[i]; u2 = d2[i];
621: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 622: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
623: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
624: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
625: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 626: d[i] = u;
627: }
628: break;
629: case 8:
1.41 noro 630: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 631: u1 = d1[i]; u2 = d2[i];
632: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 633: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
634: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
635: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 636: d[i] = u;
637: }
638: break;
639: case 16:
1.41 noro 640: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 641: u1 = d1[i]; u2 = d2[i];
642: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 643: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 644: d[i] = u;
645: }
646: break;
647: case 32:
1.41 noro 648: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 649: u1 = d1[i]; u2 = d2[i];
650: d[i] = u1>u2?u1:u2;
651: }
652: break;
653: default:
1.41 noro 654: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 655: u1 = d1[i]; u2 = d2[i];
656: for ( j = 0, u = 0; j < nd_epw; j++ ) {
657: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
658: }
659: d[i] = u;
660: }
661: break;
662: }
1.65 noro 663: #else
664: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
665: u1 = d1[i]; u2 = d2[i];
666: for ( j = 0, u = 0; j < nd_epw; j++ ) {
667: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
668: }
669: d[i] = u;
670: }
671: #endif
1.39 noro 672: TD(d) = ndl_weight(d);
1.61 noro 673: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 674: }
675:
1.61 noro 676: int ndl_weight(UINT *d)
1.1 noro 677: {
1.61 noro 678: UINT t,u;
1.1 noro 679: int i,j;
680:
1.60 noro 681: if ( current_dl_weight_vector )
682: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
683: u = GET_EXP(d,i);
684: t += MUL_WEIGHT(u,i);
685: }
686: else
687: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
688: u = d[i];
689: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
690: t += (u&nd_mask0);
691: }
1.1 noro 692: return t;
693: }
694:
1.61 noro 695: void ndl_weight_mask(UINT *d)
1.43 noro 696: {
1.61 noro 697: UINT t,u;
698: UINT *mask;
699: int i,j,k,l;
1.43 noro 700:
1.61 noro 701: l = nd_blockmask->n;
702: for ( k = 0; k < l; k++ ) {
703: mask = nd_blockmask->mask[k];
704: if ( current_dl_weight_vector )
705: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
706: u = GET_EXP_MASK(d,i,mask);
707: t += MUL_WEIGHT(u,i);
708: }
709: else
710: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
711: u = d[i]&mask[i];
712: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
713: t += (u&nd_mask0);
714: }
715: d[k+1] = t;
716: }
1.43 noro 717: }
718:
1.61 noro 719: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 720: {
721: int i;
722:
1.41 noro 723: d1 += nd_exporigin;
724: d2 += nd_exporigin;
725: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 726: if ( *d1 > *d2 )
1.32 noro 727: return nd_isrlex ? -1 : 1;
1.1 noro 728: else if ( *d1 < *d2 )
1.32 noro 729: return nd_isrlex ? 1 : -1;
1.1 noro 730: return 0;
731: }
732:
1.61 noro 733: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 734: {
735: int i,l,j,ord_o,ord_l;
736: struct order_pair *op;
1.61 noro 737: UINT t1,t2,m;
738: UINT *mask;
1.43 noro 739:
740: l = nd_blockmask->n;
741: op = nd_blockmask->order_pair;
742: for ( j = 0; j < l; j++ ) {
743: mask = nd_blockmask->mask[j];
744: ord_o = op[j].order;
745: if ( ord_o < 2 )
1.44 noro 746: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
747: else if ( t1 < t2 ) return -1;
1.43 noro 748: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 749: m = mask[i];
750: t1 = d1[i]&m;
751: t2 = d2[i]&m;
1.43 noro 752: if ( t1 > t2 )
753: return !ord_o ? -1 : 1;
754: else if ( t1 < t2 )
755: return !ord_o ? 1 : -1;
756: }
757: }
758: return 0;
759: }
760:
1.58 noro 761: /* TDH -> WW -> TD-> RL */
762:
1.61 noro 763: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 764: {
765: int i,m,e1,e2;
766:
767: if ( TD(d1) > TD(d2) ) return 1;
768: else if ( TD(d1) < TD(d2) ) return -1;
769: m = nd_nvar>>1;
770: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
771: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
772: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
773: }
774: if ( e1 > e2 ) return 1;
775: else if ( e1 < e2 ) return -1;
776: return ndl_lex_compare(d1,d2);
777: }
778:
1.61 noro 779: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 780: {
781: int i;
782:
1.41 noro 783: for ( i = 0; i < nd_wpd; i++ )
1.34 noro 784: if ( *d1++ != *d2++ )
1.1 noro 785: return 0;
786: return 1;
787: }
788:
1.61 noro 789: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 790: {
791: int i;
792:
793: switch ( nd_wpd ) {
1.41 noro 794: case 2:
1.34 noro 795: TD(d2) = TD(d1);
796: d2[1] = d1[1];
1.6 noro 797: break;
1.41 noro 798: case 3:
1.34 noro 799: TD(d2) = TD(d1);
1.6 noro 800: d2[1] = d1[1];
1.34 noro 801: d2[2] = d1[2];
1.6 noro 802: break;
803: default:
1.41 noro 804: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 805: d2[i] = d1[i];
806: break;
807: }
808: }
809:
1.61 noro 810: INLINE void ndl_zero(UINT *d)
811: {
812: int i;
813: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
814: }
815:
816: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 817: {
818: int i;
819:
1.43 noro 820: #if 1
1.6 noro 821: switch ( nd_wpd ) {
1.41 noro 822: case 2:
823: TD(d) = TD(d1)+TD(d2);
1.34 noro 824: d[1] = d1[1]+d2[1];
1.6 noro 825: break;
1.41 noro 826: case 3:
827: TD(d) = TD(d1)+TD(d2);
1.6 noro 828: d[1] = d1[1]+d2[1];
1.34 noro 829: d[2] = d1[2]+d2[2];
1.6 noro 830: break;
831: default:
1.43 noro 832: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 833: break;
834: }
1.43 noro 835: #else
836: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
837: #endif
1.6 noro 838: }
839:
1.55 noro 840: /* d1 += d2 */
1.61 noro 841: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 842: {
843: int i;
844:
845: #if 1
846: switch ( nd_wpd ) {
847: case 2:
848: TD(d1) += TD(d2);
849: d1[1] += d2[1];
850: break;
851: case 3:
852: TD(d1) += TD(d2);
853: d1[1] += d2[1];
854: d1[2] += d2[2];
855: break;
856: default:
857: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
858: break;
859: }
860: #else
861: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
862: #endif
863: }
864:
1.61 noro 865: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 866: {
867: int i;
868:
1.43 noro 869: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 870: }
871:
1.61 noro 872: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 873: {
1.61 noro 874: UINT t1,t2,u,u1,u2;
1.1 noro 875: int i,j;
876:
1.65 noro 877: #if USE_UNROLL
1.1 noro 878: switch ( nd_bpe ) {
1.62 noro 879: case 3:
880: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
881: u1 = d1[i]; u2 = d2[i];
882: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
883: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
884: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
885: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
886: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
887: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
888: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
889: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
890: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
891: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
892: }
893: return 1;
894: break;
1.1 noro 895: case 4:
1.41 noro 896: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 897: u1 = d1[i]; u2 = d2[i];
898: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 899: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
900: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
901: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
902: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
903: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
904: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
905: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 906: }
907: return 1;
908: break;
909: case 6:
1.41 noro 910: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 911: u1 = d1[i]; u2 = d2[i];
912: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 913: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
914: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
915: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
916: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 917: }
918: return 1;
919: break;
920: case 8:
1.41 noro 921: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 922: u1 = d1[i]; u2 = d2[i];
923: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 924: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
925: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
926: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 927: }
928: return 1;
929: break;
930: case 16:
1.41 noro 931: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 932: u1 = d1[i]; u2 = d2[i];
933: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 934: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 935: }
936: return 1;
937: break;
938: case 32:
1.41 noro 939: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 940: if ( d1[i] && d2[i] ) return 0;
941: return 1;
942: break;
943: default:
1.41 noro 944: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 945: u1 = d1[i]; u2 = d2[i];
946: for ( j = 0; j < nd_epw; j++ ) {
947: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
948: u1 >>= nd_bpe; u2 >>= nd_bpe;
949: }
950: }
951: return 1;
952: break;
953: }
1.65 noro 954: #else
955: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
956: u1 = d1[i]; u2 = d2[i];
957: for ( j = 0; j < nd_epw; j++ ) {
958: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
959: u1 >>= nd_bpe; u2 >>= nd_bpe;
960: }
961: }
962: return 1;
963: #endif
1.1 noro 964: }
965:
1.61 noro 966: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 967: {
1.61 noro 968: UINT u2;
969: UINT *d1;
1.5 noro 970: int i,j,ind,k;
1.1 noro 971:
1.5 noro 972: d1 = nd_bound[index];
973: ind = 0;
1.65 noro 974: #if USE_UNROLL
1.5 noro 975: switch ( nd_bpe ) {
1.62 noro 976: case 3:
977: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
978: u2 = d2[i];
979: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
980: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
981: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
982: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
983: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
984: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
985: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
986: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
987: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
988: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
989: }
990: return 0;
991: break;
1.5 noro 992: case 4:
1.41 noro 993: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 994: u2 = d2[i];
995: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
996: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
997: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
998: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
999: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
1000: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
1001: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
1002: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
1003: }
1004: return 0;
1005: break;
1006: case 6:
1.41 noro 1007: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1008: u2 = d2[i];
1009: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1010: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1011: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1012: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1013: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1014: }
1015: return 0;
1016: break;
1017: case 8:
1.41 noro 1018: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1019: u2 = d2[i];
1020: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1021: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1022: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1023: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1024: }
1025: return 0;
1026: break;
1027: case 16:
1.41 noro 1028: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1029: u2 = d2[i];
1030: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1031: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1032: }
1033: return 0;
1034: break;
1035: case 32:
1.41 noro 1036: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1037: if ( d1[i]+d2[i]<d1[i] ) return 1;
1038: return 0;
1039: break;
1040: default:
1.41 noro 1041: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1042: u2 = d2[i];
1043: k = (nd_epw-1)*nd_bpe;
1044: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1045: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1046: }
1047: return 0;
1048: break;
1049: }
1.65 noro 1050: #else
1051: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1052: u2 = d2[i];
1053: k = (nd_epw-1)*nd_bpe;
1054: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1055: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1056: }
1057: return 0;
1058: #endif
1.1 noro 1059: }
1060:
1.61 noro 1061: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1062: {
1063: int i;
1064: int r;
1065:
1.34 noro 1066: r = 0;
1.41 noro 1067: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1068: r = ((r<<16)+d[i])%REDTAB_LEN;
1069: return r;
1070: }
1071:
1.63 noro 1072: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1073: {
1.13 noro 1074: RHist r;
1.6 noro 1075: int d,k,i;
1.1 noro 1076:
1.63 noro 1077: d = ndl_hash_value(dg);
1.13 noro 1078: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1079: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1080: if ( k > 0 ) nd_notfirst++;
1081: nd_found++;
1.13 noro 1082: return r->index;
1.1 noro 1083: }
1084: }
1.13 noro 1085: if ( Reverse )
1086: for ( i = nd_psn-1; i >= 0; i-- ) {
1087: r = nd_psh[i];
1.34 noro 1088: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1089: nd_create++;
1.34 noro 1090: nd_append_red(dg,i);
1.13 noro 1091: return i;
1092: }
1093: }
1094: else
1095: for ( i = 0; i < nd_psn; i++ ) {
1096: r = nd_psh[i];
1.34 noro 1097: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1098: nd_create++;
1.34 noro 1099: nd_append_red(dg,i);
1.13 noro 1100: return i;
1101: }
1.1 noro 1102: }
1.6 noro 1103: return -1;
1.1 noro 1104: }
1105:
1.63 noro 1106: ND nd_merge(ND p1,ND p2)
1107: {
1108: int n,c;
1109: int t,can,td1,td2;
1110: ND r;
1111: NM m1,m2,mr0,mr,s;
1112:
1113: if ( !p1 ) return p2;
1114: else if ( !p2 ) return p1;
1115: else {
1116: can = 0;
1117: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1118: c = DL_COMPARE(DL(m1),DL(m2));
1119: switch ( c ) {
1120: case 0:
1121: s = m1; m1 = NEXT(m1);
1122: can++; NEXTNM2(mr0,mr,s);
1123: s = m2; m2 = NEXT(m2); FREENM(s);
1124: break;
1125: case 1:
1126: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1127: break;
1128: case -1:
1129: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1130: break;
1131: }
1132: }
1133: if ( !mr0 )
1134: if ( m1 ) mr0 = m1;
1135: else if ( m2 ) mr0 = m2;
1136: else return 0;
1137: else if ( m1 ) NEXT(mr) = m1;
1138: else if ( m2 ) NEXT(mr) = m2;
1139: else NEXT(mr) = 0;
1140: BDY(p1) = mr0;
1141: SG(p1) = MAX(SG(p1),SG(p2));
1142: LEN(p1) = LEN(p1)+LEN(p2)-can;
1143: FREEND(p2);
1144: return p1;
1145: }
1146: }
1147:
1.31 noro 1148: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1149: {
1150: int n,c;
1.34 noro 1151: int t,can,td1,td2;
1.1 noro 1152: ND r;
1153: NM m1,m2,mr0,mr,s;
1154:
1.34 noro 1155: if ( !p1 ) return p2;
1156: else if ( !p2 ) return p1;
1.71 noro 1157: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 1158: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1159: else {
1.30 noro 1160: can = 0;
1.1 noro 1161: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1162: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1163: switch ( c ) {
1164: case 0:
1.19 noro 1165: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1166: if ( t < 0 ) t += mod;
1.1 noro 1167: s = m1; m1 = NEXT(m1);
1168: if ( t ) {
1.34 noro 1169: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1170: } else {
1.34 noro 1171: can += 2; FREENM(s);
1.1 noro 1172: }
1173: s = m2; m2 = NEXT(m2); FREENM(s);
1174: break;
1175: case 1:
1176: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1177: break;
1178: case -1:
1179: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1180: break;
1181: }
1182: }
1183: if ( !mr0 )
1.34 noro 1184: if ( m1 ) mr0 = m1;
1185: else if ( m2 ) mr0 = m2;
1186: else return 0;
1187: else if ( m1 ) NEXT(mr) = m1;
1188: else if ( m2 ) NEXT(mr) = m2;
1189: else NEXT(mr) = 0;
1.1 noro 1190: BDY(p1) = mr0;
1.14 noro 1191: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1192: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1193: FREEND(p2);
1194: return p1;
1195: }
1196: }
1197:
1.31 noro 1198: ND nd_add_q(ND p1,ND p2)
1.17 noro 1199: {
1.30 noro 1200: int n,c,can;
1.17 noro 1201: ND r;
1202: NM m1,m2,mr0,mr,s;
1203: Q t;
1204:
1.34 noro 1205: if ( !p1 ) return p2;
1206: else if ( !p2 ) return p1;
1.31 noro 1207: else {
1.30 noro 1208: can = 0;
1.17 noro 1209: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1210: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1211: switch ( c ) {
1212: case 0:
1213: addq(CQ(m1),CQ(m2),&t);
1214: s = m1; m1 = NEXT(m1);
1215: if ( t ) {
1.34 noro 1216: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1217: } else {
1.34 noro 1218: can += 2; FREENM(s);
1.17 noro 1219: }
1220: s = m2; m2 = NEXT(m2); FREENM(s);
1221: break;
1222: case 1:
1223: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1224: break;
1225: case -1:
1226: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1227: break;
1228: }
1229: }
1230: if ( !mr0 )
1.34 noro 1231: if ( m1 ) mr0 = m1;
1232: else if ( m2 ) mr0 = m2;
1233: else return 0;
1234: else if ( m1 ) NEXT(mr) = m1;
1235: else if ( m2 ) NEXT(mr) = m2;
1236: else NEXT(mr) = 0;
1.17 noro 1237: BDY(p1) = mr0;
1238: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1239: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1240: FREEND(p2);
1241: return p1;
1242: }
1243: }
1244:
1.71 noro 1245: ND nd_add_sf(ND p1,ND p2)
1246: {
1247: int n,c,can;
1248: ND r;
1249: NM m1,m2,mr0,mr,s;
1250: int t;
1251:
1252: if ( !p1 ) return p2;
1253: else if ( !p2 ) return p1;
1254: else {
1255: can = 0;
1256: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1257: c = DL_COMPARE(DL(m1),DL(m2));
1258: switch ( c ) {
1259: case 0:
1260: t = _addsf(CM(m1),CM(m2));
1261: s = m1; m1 = NEXT(m1);
1262: if ( t ) {
1263: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1264: } else {
1265: can += 2; FREENM(s);
1266: }
1267: s = m2; m2 = NEXT(m2); FREENM(s);
1268: break;
1269: case 1:
1270: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1271: break;
1272: case -1:
1273: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1274: break;
1275: }
1276: }
1277: if ( !mr0 )
1278: if ( m1 ) mr0 = m1;
1279: else if ( m2 ) mr0 = m2;
1280: else return 0;
1281: else if ( m1 ) NEXT(mr) = m1;
1282: else if ( m2 ) NEXT(mr) = m2;
1283: else NEXT(mr) = 0;
1284: BDY(p1) = mr0;
1285: SG(p1) = MAX(SG(p1),SG(p2));
1286: LEN(p1) = LEN(p1)+LEN(p2)-can;
1287: FREEND(p2);
1288: return p1;
1289: }
1290: }
1291:
1.1 noro 1292: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1293: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1294: {
1.11 noro 1295: ND d;
1.1 noro 1296: NM m,mrd,tail;
1.7 noro 1297: NM mul;
1.10 noro 1298: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1299: int c,c1,c2,dummy;
1.17 noro 1300: RHist h;
1.11 noro 1301: NDV p,red;
1.69 noro 1302: Q cg,cred,gcd,tq,qq;
1.21 noro 1303: double hmag;
1.1 noro 1304:
1.69 noro 1305: if ( dn ) {
1306: if ( mod )
1307: dn->m = 1;
1308: else
1309: dn->z = ONE;
1310: }
1.1 noro 1311: if ( !g ) {
1312: *rp = 0;
1313: return 1;
1314: }
1.34 noro 1315: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1316:
1.14 noro 1317: sugar0 = sugar = SG(g);
1.1 noro 1318: n = NV(g);
1.61 noro 1319: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1320: for ( d = 0; g; ) {
1.63 noro 1321: index = ndl_find_reducer(HDL(g));
1.6 noro 1322: if ( index >= 0 ) {
1.17 noro 1323: h = nd_psh[index];
1324: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1325: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1326: nd_free(g); nd_free(d);
1327: return 0;
1328: }
1.53 noro 1329: p = ps[index];
1.71 noro 1330: if ( mod == -1 )
1331: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1332: else if ( mod ) {
1.19 noro 1333: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1334: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1335: } else {
1.17 noro 1336: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1337: chsgnq(cg,&CQ(mul));
1.20 noro 1338: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1339: if ( dn ) {
1340: mulq(dn->z,cred,&tq); dn->z = tq;
1341: }
1.16 noro 1342: }
1.55 noro 1343: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1344: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1345: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1346: tq = HCQ(g);
1.21 noro 1347: nd_removecont2(d,g);
1.69 noro 1348: if ( dn ) {
1349: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1350: }
1.21 noro 1351: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1352: }
1.1 noro 1353: } else if ( !full ) {
1354: *rp = g;
1355: return 1;
1356: } else {
1357: m = BDY(g);
1358: if ( NEXT(m) ) {
1.34 noro 1359: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1360: } else {
1361: FREEND(g); g = 0;
1362: }
1363: if ( d ) {
1.34 noro 1364: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1365: } else {
1.34 noro 1366: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1367: }
1368: }
1369: }
1.34 noro 1370: if ( d ) SG(d) = sugar;
1.1 noro 1371: *rp = d;
1372: return 1;
1373: }
1.28 noro 1374:
1.53 noro 1375: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1376: {
1377: int hindex,index;
1378: NDV p;
1379: ND u,d,red;
1380: NODE l;
1.31 noro 1381: NM mul,m,mrd,tail;
1.25 noro 1382: int sugar,psugar,n,h_reducible;
1383: PGeoBucket bucket;
1384: int c,c1,c2;
1.26 noro 1385: Q cg,cred,gcd,zzz;
1.25 noro 1386: RHist h;
1.28 noro 1387: double hmag,gmag;
1.25 noro 1388:
1389: if ( !g ) {
1390: *rp = 0;
1391: return 1;
1392: }
1393: sugar = SG(g);
1394: n = NV(g);
1.34 noro 1395: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1396: bucket = create_pbucket();
1.31 noro 1397: add_pbucket(mod,bucket,g);
1.25 noro 1398: d = 0;
1.61 noro 1399: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1400: while ( 1 ) {
1.26 noro 1401: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1402: if ( hindex < 0 ) {
1.34 noro 1403: if ( d ) SG(d) = sugar;
1.25 noro 1404: *rp = d;
1405: return 1;
1406: }
1407: g = bucket->body[hindex];
1.63 noro 1408: index = ndl_find_reducer(HDL(g));
1.25 noro 1409: if ( index >= 0 ) {
1410: h = nd_psh[index];
1411: ndl_sub(HDL(g),DL(h),DL(mul));
1412: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1413: nd_free(d);
1.25 noro 1414: free_pbucket(bucket);
1415: *rp = 0;
1416: return 0;
1417: }
1.53 noro 1418: p = ps[index];
1.71 noro 1419: if ( mod == -1 )
1420: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1421: else if ( mod ) {
1.25 noro 1422: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1423: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1424: } else {
1425: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1426: chsgnq(cg,&CQ(mul));
1.26 noro 1427: nd_mul_c_q(d,cred);
1428: mulq_pbucket(bucket,cred);
1429: g = bucket->body[hindex];
1.28 noro 1430: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1431: }
1.55 noro 1432: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1433: bucket->body[hindex] = nd_remove_head(g);
1434: red = nd_remove_head(red);
1.31 noro 1435: add_pbucket(mod,bucket,red);
1.34 noro 1436: psugar = SG(p)+TD(DL(mul));
1437: sugar = MAX(sugar,psugar);
1.28 noro 1438: if ( !mod && hmag && (gmag > hmag) ) {
1439: g = normalize_pbucket(mod,bucket);
1440: if ( !g ) {
1.34 noro 1441: if ( d ) SG(d) = sugar;
1.28 noro 1442: *rp = d;
1443: return 1;
1444: }
1445: nd_removecont2(d,g);
1446: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1447: add_pbucket(mod,bucket,g);
1.28 noro 1448: }
1.25 noro 1449: } else if ( !full ) {
1450: g = normalize_pbucket(mod,bucket);
1.34 noro 1451: if ( g ) SG(g) = sugar;
1.25 noro 1452: *rp = g;
1453: return 1;
1454: } else {
1455: m = BDY(g);
1456: if ( NEXT(m) ) {
1.34 noro 1457: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1458: } else {
1459: FREEND(g); g = 0;
1460: }
1461: bucket->body[hindex] = g;
1462: NEXT(m) = 0;
1463: if ( d ) {
1.34 noro 1464: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1465: } else {
1.34 noro 1466: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1467: }
1468: }
1469: }
1470: }
1.27 noro 1471:
1.61 noro 1472: /* input : list of NDV, cand : list of NDV */
1.28 noro 1473:
1.61 noro 1474: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1475: {
1476: int n,i,stat;
1477: ND nf,d;
1.61 noro 1478: NDV r;
1.45 noro 1479: NODE t,s;
1.69 noro 1480: union oNDC dn;
1.45 noro 1481:
1.61 noro 1482: ndv_setup(0,0,cand);
1.31 noro 1483: n = length(cand);
1.28 noro 1484:
1485: /* membercheck : list is a subset of Id(cand) ? */
1486: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1487: again:
1.61 noro 1488: if ( nd_bpe > obpe )
1489: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1490: else
1491: r = (NDV)BDY(t);
1492: d = ndvtond(0,r);
1.69 noro 1493: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1494: if ( !stat ) {
1495: nd_reconstruct(0,0,0);
1496: goto again;
1497: } else if ( nf ) return 0;
1.71 noro 1498: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1499: }
1.71 noro 1500: if ( DP_Print ) { printf("\n"); }
1.28 noro 1501: /* gbcheck : cand is a GB of Id(cand) ? */
1.34 noro 1502: if ( !nd_gb(0,1) ) return 0;
1.28 noro 1503: /* XXX */
1.23 noro 1504: return 1;
1505: }
1.1 noro 1506:
1507: ND nd_remove_head(ND p)
1508: {
1509: NM m;
1510:
1511: m = BDY(p);
1512: if ( !NEXT(m) ) {
1.34 noro 1513: FREEND(p); p = 0;
1.31 noro 1514: } else {
1.34 noro 1515: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1516: }
1.1 noro 1517: FREENM(m);
1518: return p;
1519: }
1520:
1.69 noro 1521: ND nd_separate_head(ND p,ND *head)
1522: {
1523: NM m,m0;
1524: ND r;
1525:
1526: m = BDY(p);
1527: if ( !NEXT(m) ) {
1528: *head = p; p = 0;
1529: } else {
1530: m0 = m;
1531: BDY(p) = NEXT(m); LEN(p)--;
1532: NEXT(m0) = 0;
1533: MKND(NV(p),m0,1,r);
1534: *head = r;
1535: }
1536: return p;
1537: }
1538:
1.1 noro 1539: PGeoBucket create_pbucket()
1540: {
1541: PGeoBucket g;
1542:
1543: g = CALLOC(1,sizeof(struct oPGeoBucket));
1544: g->m = -1;
1545: return g;
1546: }
1547:
1.25 noro 1548: void free_pbucket(PGeoBucket b) {
1549: int i;
1550:
1.26 noro 1551: for ( i = 0; i <= b->m; i++ )
1.25 noro 1552: if ( b->body[i] ) {
1553: nd_free(b->body[i]);
1554: b->body[i] = 0;
1555: }
1556: GC_free(b);
1557: }
1558:
1.63 noro 1559: void add_pbucket_symbolic(PGeoBucket g,ND d)
1560: {
1561: int l,i,k,m;
1562:
1563: if ( !d )
1564: return;
1565: l = LEN(d);
1566: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1567: /* 2^(k-1) < l <= 2^k (=m) */
1568: d = nd_merge(g->body[k],d);
1569: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1570: g->body[k] = 0;
1571: d = nd_merge(g->body[k+1],d);
1572: }
1573: g->body[k] = d;
1574: g->m = MAX(g->m,k);
1575: }
1576:
1.31 noro 1577: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1578: {
1.31 noro 1579: int l,i,k,m;
1.1 noro 1580:
1.31 noro 1581: if ( !d )
1582: return;
1583: l = LEN(d);
1.29 noro 1584: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1585: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1586: d = nd_add(mod,g->body[k],d);
1587: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1588: g->body[k] = 0;
1.31 noro 1589: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1590: }
1591: g->body[k] = d;
1592: g->m = MAX(g->m,k);
1593: }
1594:
1.26 noro 1595: void mulq_pbucket(PGeoBucket g,Q c)
1596: {
1597: int k;
1598:
1599: for ( k = 0; k <= g->m; k++ )
1600: nd_mul_c_q(g->body[k],c);
1601: }
1602:
1.63 noro 1603: NM remove_head_pbucket_symbolic(PGeoBucket g)
1604: {
1605: int j,i,k,c;
1606: NM head;
1607:
1608: k = g->m;
1609: j = -1;
1610: for ( i = 0; i <= k; i++ ) {
1611: if ( !g->body[i] ) continue;
1612: if ( j < 0 ) j = i;
1613: else {
1614: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1615: if ( c > 0 )
1616: j = i;
1617: else if ( c == 0 )
1618: g->body[i] = nd_remove_head(g->body[i]);
1619: }
1620: }
1621: if ( j < 0 ) return 0;
1622: else {
1623: head = BDY(g->body[j]);
1624: if ( !NEXT(head) ) {
1625: FREEND(g->body[j]);
1626: g->body[j] = 0;
1627: } else {
1628: BDY(g->body[j]) = NEXT(head);
1629: LEN(g->body[j])--;
1630: }
1631: return head;
1632: }
1633: }
1634:
1.19 noro 1635: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1636: {
1637: int j,i,c,k,nv,sum;
1.61 noro 1638: UINT *di,*dj;
1.1 noro 1639: ND gi,gj;
1640:
1641: k = g->m;
1642: while ( 1 ) {
1643: j = -1;
1644: for ( i = 0; i <= k; i++ ) {
1645: if ( !(gi = g->body[i]) )
1646: continue;
1647: if ( j < 0 ) {
1648: j = i;
1649: gj = g->body[j];
1650: dj = HDL(gj);
1.14 noro 1651: sum = HCM(gj);
1.1 noro 1652: } else {
1.34 noro 1653: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1654: if ( c > 0 ) {
1.34 noro 1655: if ( sum ) HCM(gj) = sum;
1656: else g->body[j] = nd_remove_head(gj);
1.1 noro 1657: j = i;
1658: gj = g->body[j];
1659: dj = HDL(gj);
1.14 noro 1660: sum = HCM(gj);
1.1 noro 1661: } else if ( c == 0 ) {
1.71 noro 1662: if ( mod == -1 )
1663: sum = _addsf(sum,HCM(gi));
1664: else {
1665: sum = sum+HCM(gi)-mod;
1666: if ( sum < 0 ) sum += mod;
1667: }
1.1 noro 1668: g->body[i] = nd_remove_head(gi);
1669: }
1670: }
1671: }
1.34 noro 1672: if ( j < 0 ) return -1;
1.1 noro 1673: else if ( sum ) {
1.14 noro 1674: HCM(gj) = sum;
1.26 noro 1675: return j;
1.31 noro 1676: } else
1.26 noro 1677: g->body[j] = nd_remove_head(gj);
1678: }
1679: }
1680:
1681: int head_pbucket_q(PGeoBucket g)
1682: {
1683: int j,i,c,k,nv;
1684: Q sum,t;
1685: ND gi,gj;
1686:
1687: k = g->m;
1688: while ( 1 ) {
1689: j = -1;
1690: for ( i = 0; i <= k; i++ ) {
1.34 noro 1691: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1692: if ( j < 0 ) {
1693: j = i;
1694: gj = g->body[j];
1695: sum = HCQ(gj);
1696: } else {
1697: nv = NV(gi);
1.34 noro 1698: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1699: if ( c > 0 ) {
1.34 noro 1700: if ( sum ) HCQ(gj) = sum;
1701: else g->body[j] = nd_remove_head(gj);
1.26 noro 1702: j = i;
1703: gj = g->body[j];
1704: sum = HCQ(gj);
1705: } else if ( c == 0 ) {
1706: addq(sum,HCQ(gi),&t);
1707: sum = t;
1708: g->body[i] = nd_remove_head(gi);
1709: }
1710: }
1711: }
1.34 noro 1712: if ( j < 0 ) return -1;
1.26 noro 1713: else if ( sum ) {
1714: HCQ(gj) = sum;
1.1 noro 1715: return j;
1.31 noro 1716: } else
1.1 noro 1717: g->body[j] = nd_remove_head(gj);
1718: }
1719: }
1720:
1.25 noro 1721: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1722: {
1.31 noro 1723: int i;
1.1 noro 1724: ND r,t;
1725:
1726: r = 0;
1.28 noro 1727: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1728: r = nd_add(mod,r,g->body[i]);
1.28 noro 1729: g->body[i] = 0;
1730: }
1731: g->m = -1;
1.1 noro 1732: return r;
1733: }
1734:
1.27 noro 1735: /* return value = 0 => input is not a GB */
1736:
1737: NODE nd_gb(int m,int checkonly)
1.1 noro 1738: {
1739: int i,nh,sugar,stat;
1.23 noro 1740: NODE r,g,t;
1.1 noro 1741: ND_pairs d;
1742: ND_pairs l;
1743: ND h,nf;
1.63 noro 1744: NDV nfv;
1.69 noro 1745: union oNDC dn;
1.1 noro 1746:
1.23 noro 1747: g = 0; d = 0;
1748: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1749: d = update_pairs(d,g,i);
1750: g = update_base(g,i);
1751: }
1752: sugar = 0;
1753: while ( d ) {
1754: again:
1755: l = nd_minp(d,&d);
1.14 noro 1756: if ( SG(l) != sugar ) {
1757: sugar = SG(l);
1.71 noro 1758: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1759: }
1.53 noro 1760: stat = nd_sp(m,0,l,&h);
1.1 noro 1761: if ( !stat ) {
1762: NEXT(l) = d; d = l;
1.20 noro 1763: d = nd_reconstruct(m,0,d);
1.1 noro 1764: goto again;
1765: }
1.41 noro 1766: #if USE_GEOBUCKET
1.69 noro 1767: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1768: #else
1.69 noro 1769: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1770: #endif
1.1 noro 1771: if ( !stat ) {
1772: NEXT(l) = d; d = l;
1.20 noro 1773: d = nd_reconstruct(m,0,d);
1.1 noro 1774: goto again;
1775: } else if ( nf ) {
1.27 noro 1776: if ( checkonly ) return 0;
1.71 noro 1777: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1778: nd_removecont(m,nf);
1779: nfv = ndtondv(m,nf); nd_free(nf);
1780: nh = ndv_newps(nfv,0);
1.1 noro 1781: d = update_pairs(d,g,nh);
1782: g = update_base(g,nh);
1783: FREENDP(l);
1784: } else {
1.71 noro 1785: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1786: FREENDP(l);
1787: }
1788: }
1.53 noro 1789: for ( t = g; t; t = NEXT(t) ) BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1790: return g;
1791: }
1792:
1.23 noro 1793: NODE nd_gb_trace(int m)
1.20 noro 1794: {
1795: int i,nh,sugar,stat;
1.23 noro 1796: NODE r,g,t;
1.20 noro 1797: ND_pairs d;
1798: ND_pairs l;
1799: ND h,nf,nfq;
1.63 noro 1800: NDV nfv,nfqv;
1.69 noro 1801: union oNDC dn;
1.20 noro 1802:
1.23 noro 1803: g = 0; d = 0;
1804: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1805: d = update_pairs(d,g,i);
1806: g = update_base(g,i);
1807: }
1808: sugar = 0;
1809: while ( d ) {
1810: again:
1811: l = nd_minp(d,&d);
1812: if ( SG(l) != sugar ) {
1813: sugar = SG(l);
1.71 noro 1814: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1815: }
1.53 noro 1816: stat = nd_sp(m,0,l,&h);
1.20 noro 1817: if ( !stat ) {
1818: NEXT(l) = d; d = l;
1819: d = nd_reconstruct(m,1,d);
1820: goto again;
1821: }
1.41 noro 1822: #if USE_GEOBUCKET
1.53 noro 1823: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1824: #else
1.69 noro 1825: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1826: #endif
1.20 noro 1827: if ( !stat ) {
1828: NEXT(l) = d; d = l;
1829: d = nd_reconstruct(m,1,d);
1830: goto again;
1831: } else if ( nf ) {
1.70 noro 1832: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1833: NEXT(l) = d; d = l;
1834: d = nd_reconstruct(m,1,d);
1835: goto again;
1836: }
1.20 noro 1837: if ( nfq ) {
1.70 noro 1838: /* m|HC(nfq) => failure */
1.63 noro 1839: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1840:
1.71 noro 1841: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1842: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1843: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1844: nh = ndv_newps(nfv,nfqv);
1.20 noro 1845: d = update_pairs(d,g,nh);
1846: g = update_base(g,nh);
1847: } else {
1.71 noro 1848: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1849: }
1850: } else {
1.71 noro 1851: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1852: }
1853: FREENDP(l);
1854: }
1.23 noro 1855: for ( t = g; t; t = NEXT(t) )
1.53 noro 1856: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1857: return g;
1858: }
1859:
1.23 noro 1860: int ndv_compare(NDV *p1,NDV *p2)
1861: {
1.34 noro 1862: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1863: }
1864:
1865: int ndv_compare_rev(NDV *p1,NDV *p2)
1866: {
1.34 noro 1867: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1868: }
1869:
1.61 noro 1870: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1871: {
1.69 noro 1872: int i,n,stat;
1873: ND nf,g,head;
1.23 noro 1874: NODE t,a0,a;
1.69 noro 1875: union oNDC dn;
1876: NDV *w;
1877: Q q;
1.23 noro 1878:
1.69 noro 1879: n = length(f);
1880: #if 0
1881: w = (NDV *)ALLOCA(n*sizeof(NDV));
1882: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1883: qsort(w,n,sizeof(NDV),
1884: (int (*)(const void *,const void *))ndv_compare);
1885: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1886: #endif
1887: ndv_setup(m,0,f);
1888: for ( i = 0; i < n; ) {
1889: g = ndvtond(m,nd_ps[i]);
1890: g = nd_separate_head(g,&head);
1891: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1892: if ( !stat )
1.69 noro 1893: nd_reconstruct(m,0,0);
1894: else {
1.71 noro 1895: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1896: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1897: nf = nd_add(m,head,nf);
1898: ndv_free(nd_ps[i]);
1.24 noro 1899: nd_removecont(m,nf);
1.69 noro 1900: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1901: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1902: i++;
1.23 noro 1903: }
1904: }
1.71 noro 1905: if ( DP_Print ) { printf("\n"); }
1.23 noro 1906: for ( a0 = 0, i = 0; i < n; i++ ) {
1907: NEXTNODE(a0,a);
1.69 noro 1908: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1909: }
1910: NEXT(a) = 0;
1911: return a0;
1912: }
1913:
1.1 noro 1914: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1915: {
1916: ND_pairs d1,nd,cur,head,prev,remove;
1917:
1918: if ( !g ) return d;
1919: d = crit_B(d,t);
1920: d1 = nd_newpairs(g,t);
1921: d1 = crit_M(d1);
1922: d1 = crit_F(d1);
1.55 noro 1923: if ( do_weyl )
1924: head = d1;
1925: else {
1926: prev = 0; cur = head = d1;
1927: while ( cur ) {
1928: if ( crit_2( cur->i1,cur->i2 ) ) {
1929: remove = cur;
1930: if ( !prev ) head = cur = NEXT(cur);
1931: else cur = NEXT(prev) = NEXT(cur);
1932: FREENDP(remove);
1933: } else {
1934: prev = cur; cur = NEXT(cur);
1935: }
1.1 noro 1936: }
1937: }
1938: if ( !d )
1939: return head;
1940: else {
1941: nd = d;
1.34 noro 1942: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1943: NEXT(nd) = head;
1944: return d;
1945: }
1946: }
1947:
1948: ND_pairs nd_newpairs( NODE g, int t )
1949: {
1950: NODE h;
1.61 noro 1951: UINT *dl;
1.34 noro 1952: int ts,s;
1.1 noro 1953: ND_pairs r,r0;
1954:
1.20 noro 1955: dl = DL(nd_psh[t]);
1.34 noro 1956: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1957: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1958: NEXTND_pairs(r0,r);
1959: r->i1 = (int)BDY(h);
1960: r->i2 = t;
1.20 noro 1961: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1962: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1963: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1964: }
1965: NEXT(r) = 0;
1966: return r0;
1967: }
1968:
1969: ND_pairs crit_B( ND_pairs d, int s )
1970: {
1971: ND_pairs cur,head,prev,remove;
1.61 noro 1972: UINT *t,*tl,*lcm;
1.1 noro 1973: int td,tdl;
1974:
1975: if ( !d ) return 0;
1.20 noro 1976: t = DL(nd_psh[s]);
1.1 noro 1977: prev = 0;
1978: head = cur = d;
1.61 noro 1979: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1980: while ( cur ) {
1981: tl = cur->lcm;
1982: if ( ndl_reducible(tl,t)
1.20 noro 1983: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1984: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1985: remove = cur;
1986: if ( !prev ) {
1987: head = cur = NEXT(cur);
1988: } else {
1989: cur = NEXT(prev) = NEXT(cur);
1990: }
1991: FREENDP(remove);
1992: } else {
1.34 noro 1993: prev = cur; cur = NEXT(cur);
1.1 noro 1994: }
1995: }
1996: return head;
1997: }
1998:
1999: ND_pairs crit_M( ND_pairs d1 )
2000: {
2001: ND_pairs e,d2,d3,dd,p;
1.61 noro 2002: UINT *id,*jd;
1.1 noro 2003:
2004: for ( dd = 0, e = d1; e; e = d3 ) {
2005: if ( !(d2 = NEXT(e)) ) {
2006: NEXT(e) = dd;
2007: return e;
2008: }
1.34 noro 2009: id = LCM(e);
1.1 noro 2010: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2011: p = NEXT(d2);
2012: jd = LCM(d2);
2013: if ( ndl_equal(jd,id) )
2014: ;
2015: else if ( TD(jd) > TD(id) )
1.1 noro 2016: if ( ndl_reducible(jd,id) ) continue;
2017: else ;
1.34 noro 2018: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2019: NEXT(d2) = d3;
2020: d3 = d2;
2021: }
2022: NEXT(e) = dd;
2023: dd = e;
2024: continue;
2025: /**/
2026: delit: NEXT(d2) = d3;
2027: d3 = d2;
2028: for ( ; p; p = d2 ) {
2029: d2 = NEXT(p);
2030: NEXT(p) = d3;
2031: d3 = p;
2032: }
2033: FREENDP(e);
2034: }
2035: return dd;
2036: }
2037:
2038: ND_pairs crit_F( ND_pairs d1 )
2039: {
2040: ND_pairs rest, head,remove;
2041: ND_pairs last, p, r, w;
2042: int s;
2043:
2044: for ( head = last = 0, p = d1; NEXT(p); ) {
2045: r = w = equivalent_pairs(p,&rest);
1.14 noro 2046: s = SG(r);
1.1 noro 2047: w = NEXT(w);
2048: while ( w ) {
2049: if ( crit_2(w->i1,w->i2) ) {
2050: r = w;
2051: w = NEXT(w);
2052: while ( w ) {
2053: remove = w;
2054: w = NEXT(w);
2055: FREENDP(remove);
2056: }
2057: break;
1.14 noro 2058: } else if ( SG(w) < s ) {
1.1 noro 2059: FREENDP(r);
2060: r = w;
1.14 noro 2061: s = SG(r);
1.1 noro 2062: w = NEXT(w);
2063: } else {
2064: remove = w;
2065: w = NEXT(w);
2066: FREENDP(remove);
2067: }
2068: }
2069: if ( last ) NEXT(last) = r;
2070: else head = r;
2071: NEXT(last = r) = 0;
2072: p = rest;
2073: if ( !p ) return head;
2074: }
2075: if ( !last ) return p;
2076: NEXT(last) = p;
2077: return head;
2078: }
2079:
2080: int crit_2( int dp1, int dp2 )
2081: {
1.20 noro 2082: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2083: }
2084:
1.40 noro 2085: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2086: {
2087: ND_pairs w,p,r,s;
1.61 noro 2088: UINT *d;
1.1 noro 2089:
2090: w = d1;
1.34 noro 2091: d = LCM(w);
1.1 noro 2092: s = NEXT(w);
2093: NEXT(w) = 0;
2094: for ( r = 0; s; s = p ) {
2095: p = NEXT(s);
1.34 noro 2096: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2097: NEXT(s) = w; w = s;
1.1 noro 2098: } else {
1.39 noro 2099: NEXT(s) = r; r = s;
1.1 noro 2100: }
2101: }
2102: *prest = r;
2103: return w;
2104: }
2105:
2106: NODE update_base(NODE nd,int ndp)
2107: {
1.61 noro 2108: UINT *dl, *dln;
1.1 noro 2109: NODE last, p, head;
2110:
1.20 noro 2111: dl = DL(nd_psh[ndp]);
1.1 noro 2112: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2113: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2114: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2115: p = NEXT(p);
2116: if ( last ) NEXT(last) = p;
2117: } else {
2118: if ( !last ) head = p;
2119: p = NEXT(last = p);
2120: }
2121: }
2122: head = append_one(head,ndp);
2123: return head;
2124: }
2125:
2126: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2127: {
2128: ND_pairs m,ml,p,l;
1.61 noro 2129: UINT *lcm;
1.33 noro 2130: int s,td,len,tlen,c,c1;
1.1 noro 2131:
2132: if ( !(p = NEXT(m = d)) ) {
2133: *prest = p;
2134: NEXT(m) = 0;
2135: return m;
2136: }
1.14 noro 2137: s = SG(m);
1.33 noro 2138: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1.34 noro 2139: if ( (SG(p) < s)
2140: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
1.39 noro 2141: ml = l; m = p; s = SG(m);
1.1 noro 2142: }
2143: if ( !ml ) *prest = NEXT(m);
2144: else {
2145: NEXT(ml) = NEXT(m);
2146: *prest = d;
2147: }
2148: NEXT(m) = 0;
2149: return m;
2150: }
2151:
1.63 noro 2152: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2153: {
1.69 noro 2154: int msugar,i;
1.63 noro 2155: ND_pairs t,dm0,dm,dr0,dr;
2156:
2157: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2158: if ( SG(t) < msugar ) msugar = SG(t);
2159: dm0 = 0; dr0 = 0;
1.69 noro 2160: for ( i = 0, t = d; t; t = NEXT(t) )
2161: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2162: if ( dm0 ) NEXT(dm) = t;
2163: else dm0 = t;
2164: dm = t;
1.69 noro 2165: i++;
1.63 noro 2166: } else {
2167: if ( dr0 ) NEXT(dr) = t;
2168: else dr0 = t;
2169: dr = t;
2170: }
2171: NEXT(dm) = 0;
2172: if ( dr0 ) NEXT(dr) = 0;
2173: *prest = dr0;
2174: return dm0;
2175: }
2176:
2177: int ndv_newps(NDV a,NDV aq)
1.1 noro 2178: {
1.3 noro 2179: int len;
1.13 noro 2180: RHist r;
1.20 noro 2181: NDV b;
1.3 noro 2182:
1.1 noro 2183: if ( nd_psn == nd_pslen ) {
2184: nd_pslen *= 2;
1.11 noro 2185: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2186: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2187: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2188: nd_bound = (UINT **)
2189: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2190: }
1.39 noro 2191: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2192: nd_ps[nd_psn] = a;
1.39 noro 2193: if ( aq ) {
1.63 noro 2194: nd_ps_trace[nd_psn] = aq;
2195: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2196: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2197: } else {
1.63 noro 2198: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2199: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2200: }
1.1 noro 2201: return nd_psn++;
2202: }
2203:
1.61 noro 2204: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2205: {
1.5 noro 2206: int i,j,td,len,max;
1.1 noro 2207: NODE s,s0,f0;
1.61 noro 2208: UINT *d;
1.13 noro 2209: RHist r;
1.69 noro 2210: NDV *w;
1.61 noro 2211: NDV a,am;
1.11 noro 2212:
2213: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2214:
1.69 noro 2215: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2216: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2217: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2218: qsort(w,nd_psn,sizeof(NDV),
2219: (int (*)(const void *,const void *))ndv_compare);
2220: nd_pslen = 2*nd_psn;
1.11 noro 2221: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2222: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2223: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2224: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2225:
1.11 noro 2226: if ( !nd_red )
1.13 noro 2227: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2228: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2229: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2230: if ( trace ) {
1.69 noro 2231: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2232: ndv_removecont(0,a);
2233: am = nd_ps[i] = ndv_dup(mod,a);
2234: ndv_mod(mod,am);
2235: ndv_removecont(mod,am);
2236: } else {
1.69 noro 2237: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2238: ndv_removecont(mod,a);
1.39 noro 2239: }
1.61 noro 2240: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2241: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2242: nd_psh[i] = r;
2243: }
2244: }
2245:
1.63 noro 2246: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2247: {
1.61 noro 2248: VL tv,fv,vv,vc;
1.1 noro 2249: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2250: int e,max,nvar;
1.69 noro 2251: NDV b;
1.1 noro 2252:
1.74 noro 2253: ndv_alloc = 0;
1.1 noro 2254: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2255: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.32 noro 2256: nd_init_ord(ord);
1.61 noro 2257: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2258: for ( tv = vv; tv; tv = NEXT(tv) ) {
2259: e = getdeg(tv->v,(P)BDY(t));
2260: max = MAX(e,max);
2261: }
2262: nd_setup_parameters(nvar,max);
1.1 noro 2263: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2264: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 2265: if ( m ) ndv_mod(m,b);
1.61 noro 2266: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2267: }
2268: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2269: ndv_setup(m,0,fd0);
1.63 noro 2270: x = f4?nd_f4(m):nd_gb(m,0);
1.61 noro 2271: x = ndv_reducebase(x);
2272: x = ndv_reduceall(m,x);
1.23 noro 2273: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2274: NEXTNODE(r0,r);
1.61 noro 2275: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2276: }
2277: if ( r0 ) NEXT(r) = 0;
2278: MKLIST(*rp,r0);
1.74 noro 2279: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2280: }
2281:
1.52 noro 2282: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2283: {
2284: struct order_spec ord1;
1.61 noro 2285: VL tv,fv,vv,vc;
1.27 noro 2286: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2287: int m,nocheck,nvar,mindex,e,max;
2288: NDV c;
2289: NMV a;
1.27 noro 2290: P p;
1.61 noro 2291: EPOS oepos;
2292: int obpe,oadv,wmax,i,len,cbpe;
1.20 noro 2293:
2294: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 noro 2295: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.52 noro 2296: nocheck = 0;
2297: mindex = 0;
2298:
2299: /* setup modulus */
2300: if ( trace < 0 ) {
2301: trace = -trace;
2302: nocheck = 1;
2303: }
2304: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2305: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2306: for ( tv = vv; tv; tv = NEXT(tv) ) {
2307: e = getdeg(tv->v,(P)BDY(t));
2308: max = MAX(e,max);
1.23 noro 2309: }
1.61 noro 2310: nd_init_ord(ord);
2311: nd_setup_parameters(nvar,max);
2312: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2313: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2314: c = ptondv(CO,vv,(P)BDY(t));
2315: if ( c ) {
2316: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2317: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2318: }
1.61 noro 2319: }
2320: if ( in0 ) NEXT(in) = 0;
2321: if ( fd0 ) NEXT(fd) = 0;
2322: if ( homo ) {
2323: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2324: c = (NDV)BDY(t); len = LEN(c);
2325: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2326: wmax = MAX(TD(DL(a)),wmax);
2327: }
2328: homogenize_order(ord,nvar,&ord1);
2329: nd_init_ord(&ord1);
2330: nd_setup_parameters(nvar+1,wmax);
2331: for ( t = fd0; t; t = NEXT(t) )
2332: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2333: }
1.52 noro 2334: while ( 1 ) {
1.61 noro 2335: ndv_setup(m,1,fd0);
1.27 noro 2336: cand = nd_gb_trace(m);
1.52 noro 2337: if ( !cand ) {
2338: /* failure */
1.61 noro 2339: if ( trace > 1 ) { *rp = 0; return; }
2340: else m = get_lprime(++mindex);
1.52 noro 2341: continue;
2342: }
1.27 noro 2343: if ( homo ) {
2344: /* dehomogenization */
1.61 noro 2345: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2346: nd_init_ord(ord);
1.61 noro 2347: nd_setup_parameters(nvar,0);
1.27 noro 2348: }
1.61 noro 2349: cand = ndv_reducebase(cand);
2350: cand = ndv_reduceall(0,cand);
2351: if ( nocheck )
2352: break;
2353: cbpe = nd_bpe;
2354: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2355: /* success */
2356: break;
2357: else if ( trace > 1 ) {
2358: /* failure */
2359: *rp = 0; return;
1.61 noro 2360: } else {
1.52 noro 2361: /* try the next modulus */
2362: m = get_lprime(++mindex);
1.61 noro 2363: /* reset the parameters */
2364: if ( homo ) {
2365: nd_init_ord(&ord1);
2366: nd_setup_parameters(nvar+1,wmax);
2367: } else {
2368: nd_init_ord(ord);
2369: nd_setup_parameters(nvar,max);
2370: }
2371: }
1.52 noro 2372: }
1.27 noro 2373: /* dp->p */
1.61 noro 2374: nd_bpe = cbpe;
1.73 noro 2375: nd_setup_parameters(nd_nvar,0);
1.61 noro 2376: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2377: MKLIST(*rp,cand);
1.1 noro 2378: }
2379:
1.61 noro 2380: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2381: {
1.61 noro 2382: UINT *d;
1.57 noro 2383: int i,j,l,s,ord_l;
1.43 noro 2384: struct order_pair *op;
1.1 noro 2385:
2386: d = dl->d;
1.41 noro 2387: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2388: if ( nd_blockmask ) {
2389: l = nd_blockmask->n;
2390: op = nd_blockmask->order_pair;
2391: for ( j = 0, s = 0; j < l; j++ ) {
2392: ord_l = op[j].length;
1.57 noro 2393: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2394: }
2395: TD(r) = ndl_weight(r);
1.61 noro 2396: ndl_weight_mask(r);
1.43 noro 2397: } else {
1.56 noro 2398: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2399: TD(r) = ndl_weight(r);
2400: }
1.1 noro 2401: }
2402:
1.61 noro 2403: DL ndltodl(int n,UINT *ndl)
1.1 noro 2404: {
2405: DL dl;
2406: int *d;
1.57 noro 2407: int i,j,l,s,ord_l;
1.43 noro 2408: struct order_pair *op;
1.1 noro 2409:
2410: NEWDL(dl,n);
1.34 noro 2411: dl->td = TD(ndl);
1.1 noro 2412: d = dl->d;
1.43 noro 2413: if ( nd_blockmask ) {
2414: l = nd_blockmask->n;
2415: op = nd_blockmask->order_pair;
2416: for ( j = 0, s = 0; j < l; j++ ) {
2417: ord_l = op[j].length;
1.57 noro 2418: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2419: }
2420: } else {
1.56 noro 2421: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2422: }
1.1 noro 2423: return dl;
2424: }
2425:
1.61 noro 2426: void ndl_print(UINT *dl)
1.1 noro 2427: {
2428: int n;
1.57 noro 2429: int i,j,l,ord_l,s,s0;
1.43 noro 2430: struct order_pair *op;
1.1 noro 2431:
2432: n = nd_nvar;
2433: printf("<<");
1.43 noro 2434: if ( nd_blockmask ) {
2435: l = nd_blockmask->n;
2436: op = nd_blockmask->order_pair;
2437: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2438: ord_l = op[j].length;
1.57 noro 2439: for ( i = 0; i < ord_l; i++, s++ )
2440: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2441: }
2442: } else {
1.56 noro 2443: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2444: }
1.1 noro 2445: printf(">>");
2446: }
2447:
2448: void nd_print(ND p)
2449: {
2450: NM m;
2451:
2452: if ( !p )
2453: printf("0\n");
2454: else {
2455: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2456: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2457: else printf("+%d*",CM(m));
1.14 noro 2458: ndl_print(DL(m));
1.1 noro 2459: }
2460: printf("\n");
2461: }
2462: }
2463:
1.16 noro 2464: void nd_print_q(ND p)
2465: {
2466: NM m;
2467:
2468: if ( !p )
2469: printf("0\n");
2470: else {
2471: for ( m = BDY(p); m; m = NEXT(m) ) {
2472: printf("+");
1.74 noro 2473: printexpr(CO,(Obj)CQ(m));
1.16 noro 2474: printf("*");
2475: ndl_print(DL(m));
2476: }
2477: printf("\n");
2478: }
2479: }
2480:
1.1 noro 2481: void ndp_print(ND_pairs d)
2482: {
2483: ND_pairs t;
2484:
1.34 noro 2485: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2486: printf("\n");
2487: }
2488:
1.20 noro 2489: void nd_removecont(int mod,ND p)
1.16 noro 2490: {
2491: int i,n;
2492: Q *w;
2493: Q dvr,t;
2494: NM m;
1.21 noro 2495: struct oVECT v;
2496: N q,r;
1.16 noro 2497:
1.71 noro 2498: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2499: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2500: else {
2501: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2502: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2503: v.len = n;
2504: v.body = (pointer *)w;
1.34 noro 2505: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2506: removecont_array(w,n);
2507: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2508: }
2509: }
2510:
1.21 noro 2511: void nd_removecont2(ND p1,ND p2)
2512: {
2513: int i,n1,n2,n;
2514: Q *w;
2515: Q dvr,t;
2516: NM m;
2517: struct oVECT v;
2518: N q,r;
2519:
2520: if ( !p1 ) {
2521: nd_removecont(0,p2); return;
2522: } else if ( !p2 ) {
2523: nd_removecont(0,p1); return;
2524: }
2525: n1 = nd_length(p1);
2526: n2 = nd_length(p2);
2527: n = n1+n2;
2528: w = (Q *)ALLOCA(n*sizeof(Q));
2529: v.len = n;
2530: v.body = (pointer *)w;
1.34 noro 2531: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2532: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2533: removecont_array(w,n);
2534: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2535: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2536: }
2537:
1.20 noro 2538: void ndv_removecont(int mod,NDV p)
1.16 noro 2539: {
2540: int i,len;
2541: Q *w;
2542: Q dvr,t;
2543: NMV m;
2544:
1.71 noro 2545: if ( mod == -1 )
2546: ndv_mul_c(mod,p,_invsf(HCM(p)));
2547: else if ( mod )
1.20 noro 2548: ndv_mul_c(mod,p,invm(HCM(p),mod));
2549: else {
2550: len = p->len;
2551: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2552: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2553: sortbynm(w,len);
2554: qltozl(w,len,&dvr);
2555: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2556: divq(CQ(m),dvr,&t); CQ(m) = t;
2557: }
1.16 noro 2558: }
1.21 noro 2559: }
2560:
1.61 noro 2561: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2562: {
2563: int len,i,max;
2564: NMV m,mr0,mr,t;
2565:
2566: len = p->len;
2567: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2568: max = MAX(max,TD(DL(m)));
2569: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2570: m = (NMV)((char *)mr0+(len-1)*oadv);
2571: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2572: t = (NMV)ALLOCA(nmv_adv);
2573: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2574: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2575: CQ(mr) = CQ(m);
2576: ndl_copy(DL(t),DL(mr));
2577: }
2578: NV(p)++;
2579: BDY(p) = mr0;
2580: }
2581:
1.45 noro 2582: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2583: {
1.45 noro 2584: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2585: Q *w;
2586: Q dvr,t;
2587: NMV m,r;
2588:
2589: len = p->len;
2590: newnvar = nd_nvar-1;
1.48 noro 2591: newexporigin = nd_get_exporigin(ord);
1.45 noro 2592: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2593: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2594: ndl_dehomogenize(DL(m));
1.23 noro 2595: if ( newwpd != nd_wpd ) {
1.61 noro 2596: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2597: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2598: CQ(r) = CQ(m);
2599: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2600: adj = nd_exporigin-newexporigin;
2601: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2602: }
2603: }
2604: NV(p)--;
2605: }
2606:
1.21 noro 2607: void removecont_array(Q *c,int n)
2608: {
2609: struct oVECT v;
2610: Q d0,d1,a,u,u1,gcd;
2611: int i;
2612: N qn,rn,gn;
2613: Q *q,*r;
2614:
2615: q = (Q *)ALLOCA(n*sizeof(Q));
2616: r = (Q *)ALLOCA(n*sizeof(Q));
2617: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2618: igcdv_estimate(&v,&d0);
2619: for ( i = 0; i < n; i++ ) {
2620: divn(NM(c[i]),NM(d0),&qn,&rn);
2621: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2622: NTOQ(rn,SGN(c[i]),r[i]);
2623: }
1.34 noro 2624: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2625: if ( i < n ) {
2626: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2627: igcdv(&v,&d1);
2628: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2629: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2630: for ( i = 0; i < n; i++ ) {
2631: mulq(a,q[i],&u);
2632: if ( r[i] ) {
2633: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2634: addq(u,u1,&q[i]);
2635: } else
2636: q[i] = u;
2637: }
2638: }
1.34 noro 2639: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2640: }
2641:
1.19 noro 2642: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2643: {
2644: NM m;
2645: int c,c1;
2646:
1.34 noro 2647: if ( !p ) return;
1.71 noro 2648: if ( mod == -1 )
2649: for ( m = BDY(p); m; m = NEXT(m) )
2650: CM(m) = _mulsf(CM(m),mul);
2651: else
2652: for ( m = BDY(p); m; m = NEXT(m) ) {
2653: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2654: }
1.1 noro 2655: }
2656:
1.16 noro 2657: void nd_mul_c_q(ND p,Q mul)
2658: {
2659: NM m;
2660: Q c;
2661:
1.34 noro 2662: if ( !p ) return;
1.16 noro 2663: for ( m = BDY(p); m; m = NEXT(m) ) {
2664: mulq(CQ(m),mul,&c); CQ(m) = c;
2665: }
2666: }
2667:
1.61 noro 2668: void nd_mul_c_p(VL vl,ND p,P mul)
2669: {
2670: NM m;
2671: P c;
2672:
2673: if ( !p ) return;
2674: for ( m = BDY(p); m; m = NEXT(m) ) {
2675: mulp(vl,CP(m),mul,&c); CP(m) = c;
2676: }
2677: }
2678:
1.1 noro 2679: void nd_free(ND p)
2680: {
2681: NM t,s;
2682:
1.34 noro 2683: if ( !p ) return;
1.1 noro 2684: t = BDY(p);
2685: while ( t ) {
2686: s = NEXT(t);
2687: FREENM(t);
2688: t = s;
2689: }
2690: FREEND(p);
2691: }
2692:
1.23 noro 2693: void ndv_free(NDV p)
2694: {
2695: GC_free(BDY(p));
2696: }
2697:
1.61 noro 2698: void nd_append_red(UINT *d,int i)
1.1 noro 2699: {
1.13 noro 2700: RHist m,m0;
1.1 noro 2701: int h;
2702:
1.13 noro 2703: NEWRHist(m);
1.34 noro 2704: h = ndl_hash_value(d);
1.13 noro 2705: m->index = i;
1.14 noro 2706: ndl_copy(d,DL(m));
1.1 noro 2707: NEXT(m) = nd_red[h];
2708: nd_red[h] = m;
2709: }
2710:
1.61 noro 2711: UINT *ndv_compute_bound(NDV p)
1.1 noro 2712: {
1.61 noro 2713: UINT *d1,*d2,*t;
2714: UINT u;
1.57 noro 2715: int i,j,k,l,len,ind;
1.45 noro 2716: NMV m;
1.1 noro 2717:
2718: if ( !p )
2719: return 0;
1.61 noro 2720: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2721: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2722: len = LEN(p);
2723: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2724: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2725: ndl_lcm(DL(m),d1,d2);
1.1 noro 2726: t = d1; d1 = d2; d2 = t;
2727: }
1.12 noro 2728: l = nd_nvar+31;
1.61 noro 2729: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2730: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2731: u = d1[i];
2732: k = (nd_epw-1)*nd_bpe;
2733: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2734: t[ind] = (u>>k)&nd_mask0;
2735: }
2736: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2737: return t;
2738: }
2739:
1.48 noro 2740: int nd_get_exporigin(struct order_spec *ord)
2741: {
1.51 noro 2742: switch ( ord->id ) {
1.41 noro 2743: case 0:
1.48 noro 2744: return 1;
1.41 noro 2745: case 1:
2746: /* block order */
1.43 noro 2747: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2748: return ord->ord.block.length+1;
1.41 noro 2749: case 2:
1.52 noro 2750: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2751: }
1.48 noro 2752: }
2753:
1.61 noro 2754: void nd_setup_parameters(int nvar,int max) {
1.57 noro 2755: int i,j,n,elen,ord_o,ord_l,l,s;
2756: struct order_pair *op;
1.73 noro 2757: int bpe;
1.48 noro 2758:
1.73 noro 2759: if ( !max ) bpe = nd_bpe;
2760: else if ( max < 2 ) bpe = 1;
2761: else if ( max < 4 ) bpe = 2;
2762: else if ( max < 8 ) bpe = 3;
2763: else if ( max < 16 ) bpe = 4;
2764: else if ( max < 32 ) bpe = 5;
2765: else if ( max < 64 ) bpe = 6;
2766: else if ( max < 256 ) bpe = 8;
2767: else if ( max < 1024 ) bpe = 10;
2768: else if ( max < 65536 ) bpe = 16;
2769: else bpe = 32;
2770: if ( bpe != nd_bpe || nvar != nd_nvar )
2771: nd_free_private_storage();
2772: nd_bpe = bpe;
2773: nd_nvar = nvar;
1.61 noro 2774: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2775: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2776:
2777: nd_exporigin = nd_get_exporigin(nd_ord);
1.43 noro 2778: nd_wpd = nd_exporigin+elen;
1.57 noro 2779:
1.1 noro 2780: if ( nd_bpe < 32 ) {
2781: nd_mask0 = (1<<nd_bpe)-1;
2782: } else {
2783: nd_mask0 = 0xffffffff;
2784: }
2785: bzero(nd_mask,sizeof(nd_mask));
2786: nd_mask1 = 0;
2787: for ( i = 0; i < nd_epw; i++ ) {
2788: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2789: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2790: }
1.61 noro 2791: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2792: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2793: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2794: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2795: }
2796:
1.20 noro 2797: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2798: {
1.37 noro 2799: int i,obpe,oadv,h;
1.71 noro 2800: static NM prev_nm_free_list;
2801: static ND_pairs prev_ndp_free_list;
1.13 noro 2802: RHist mr0,mr;
2803: RHist r;
1.37 noro 2804: RHist *old_red;
1.71 noro 2805: ND_pairs s0,s,t;
1.43 noro 2806: EPOS oepos;
1.15 noro 2807:
1.1 noro 2808: obpe = nd_bpe;
1.11 noro 2809: oadv = nmv_adv;
1.43 noro 2810: oepos = nd_epos;
1.63 noro 2811: if ( obpe < 2 ) nd_bpe = 2;
2812: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2813: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2814: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2815: else if ( obpe < 6 ) nd_bpe = 6;
2816: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2817: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2818: else if ( obpe < 16 ) nd_bpe = 16;
2819: else if ( obpe < 32 ) nd_bpe = 32;
2820: else error("nd_reconstruct : exponent too large");
1.5 noro 2821:
1.73 noro 2822: nd_setup_parameters(nd_nvar,0);
1.1 noro 2823: prev_nm_free_list = _nm_free_list;
2824: prev_ndp_free_list = _ndp_free_list;
2825: _nm_free_list = 0;
2826: _ndp_free_list = 0;
1.53 noro 2827: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2828: if ( trace )
2829: for ( i = nd_psn-1; i >= 0; i-- )
2830: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2831: s0 = 0;
2832: for ( t = d; t; t = NEXT(t) ) {
2833: NEXTND_pairs(s0,s);
2834: s->i1 = t->i1;
2835: s->i2 = t->i2;
1.14 noro 2836: SG(s) = SG(t);
1.61 noro 2837: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2838: }
1.37 noro 2839:
2840: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2841: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2842: old_red[i] = nd_red[i];
2843: nd_red[i] = 0;
2844: }
2845: for ( i = 0; i < REDTAB_LEN; i++ )
2846: for ( r = old_red[i]; r; r = NEXT(r) ) {
2847: NEWRHist(mr);
1.13 noro 2848: mr->index = r->index;
1.20 noro 2849: SG(mr) = SG(r);
1.61 noro 2850: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2851: h = ndl_hash_value(DL(mr));
2852: NEXT(mr) = nd_red[h];
2853: nd_red[h] = mr;
1.6 noro 2854: }
1.37 noro 2855: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2856: old_red = 0;
1.11 noro 2857: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2858: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2859: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2860: nd_psh[i] = r;
1.11 noro 2861: }
1.1 noro 2862: if ( s0 ) NEXT(s) = 0;
2863: prev_nm_free_list = 0;
2864: prev_ndp_free_list = 0;
1.71 noro 2865: #if 0
1.1 noro 2866: GC_gcollect();
1.71 noro 2867: #endif
1.1 noro 2868: return s0;
2869: }
2870:
1.61 noro 2871: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2872: {
1.57 noro 2873: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2874: struct order_pair *op;
1.1 noro 2875:
2876: n = nd_nvar;
1.61 noro 2877: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2878: omask0 = (1<<obpe)-1;
1.34 noro 2879: TD(r) = TD(d);
1.41 noro 2880: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2881: if ( nd_blockmask ) {
2882: l = nd_blockmask->n;
2883: op = nd_blockmask->order_pair;
2884: for ( i = 1; i < nd_exporigin; i++ )
2885: r[i] = d[i];
2886: for ( j = 0, s = 0; j < l; j++ ) {
2887: ord_l = op[j].length;
1.57 noro 2888: for ( i = 0; i < ord_l; i++, s++ ) {
2889: ei = GET_EXP_OLD(d,s);
2890: PUT_EXP(r,s,ei);
2891: }
1.1 noro 2892: }
1.43 noro 2893: } else {
1.56 noro 2894: for ( i = 0; i < n; i++ ) {
2895: ei = GET_EXP_OLD(d,i);
2896: PUT_EXP(r,i,ei);
2897: }
1.1 noro 2898: }
2899: }
1.3 noro 2900:
1.6 noro 2901: ND nd_copy(ND p)
2902: {
2903: NM m,mr,mr0;
1.41 noro 2904: int c,n;
1.6 noro 2905: ND r;
2906:
2907: if ( !p )
2908: return 0;
2909: else {
2910: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2911: NEXTNM(mr0,mr);
1.14 noro 2912: CM(mr) = CM(m);
2913: ndl_copy(DL(m),DL(mr));
1.6 noro 2914: }
2915: NEXT(mr) = 0;
1.31 noro 2916: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2917: SG(r) = SG(p);
1.6 noro 2918: return r;
2919: }
2920: }
2921:
1.53 noro 2922: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2923: {
2924: NM m;
2925: NDV p1,p2;
2926: ND t1,t2;
1.61 noro 2927: UINT *lcm;
1.31 noro 2928: int td;
1.11 noro 2929:
1.53 noro 2930: if ( trace ) {
2931: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
2932: } else {
1.20 noro 2933: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
2934: }
1.34 noro 2935: lcm = LCM(p);
1.11 noro 2936: NEWNM(m);
1.20 noro 2937: CQ(m) = HCQ(p2);
1.34 noro 2938: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 2939: if ( ndl_check_bound2(p->i1,DL(m)) )
2940: return 0;
1.55 noro 2941: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 2942: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
2943: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 2944: else chsgnq(HCQ(p1),&CQ(m));
2945: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 2946: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 2947: nd_free(t1);
2948: return 0;
2949: }
1.55 noro 2950: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 2951: *rp = nd_add(mod,t1,t2);
1.11 noro 2952: FREENM(m);
2953: return 1;
2954: }
2955:
1.19 noro 2956: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 2957: {
2958: NMV m;
2959: int c,c1,len,i;
2960:
1.34 noro 2961: if ( !p ) return;
1.14 noro 2962: len = LEN(p);
1.71 noro 2963: if ( mod == -1 )
2964: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
2965: CM(m) = _mulsf(CM(m),mul);
2966: else
2967: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2968: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2969: }
1.11 noro 2970: }
2971:
1.16 noro 2972: void ndv_mul_c_q(NDV p,Q mul)
2973: {
2974: NMV m;
2975: Q c;
2976: int len,i;
2977:
1.34 noro 2978: if ( !p ) return;
1.16 noro 2979: len = LEN(p);
2980: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2981: mulq(CQ(m),mul,&c); CQ(m) = c;
2982: }
2983: }
2984:
1.55 noro 2985: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
2986: int n2,i,j,l,n,tlen;
1.61 noro 2987: UINT *d0;
1.55 noro 2988: NM *tab,*psum;
2989: ND s,r;
2990: NM t;
2991: NMV m1;
2992:
2993: if ( !p ) return 0;
2994: n = NV(p); n2 = n>>1;
2995: d0 = DL(m0);
2996: l = LEN(p);
2997: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
2998: tab = (NM *)ALLOCA(tlen*sizeof(NM));
2999: psum = (NM *)ALLOCA(tlen*sizeof(NM));
3000: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 3001: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
3002: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 3003: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3004: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3005: for ( j = 0; j < tlen; j++ ) {
3006: if ( tab[j] ) {
3007: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3008: }
3009: }
3010: }
3011: for ( i = tlen-1, r = 0; i >= 0; i-- )
3012: if ( psum[i] ) {
3013: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3014: MKND(n,psum[i],j,s);
3015: r = nd_add(mod,r,s);
3016: }
1.56 noro 3017: if ( r ) SG(r) = SG(p)+TD(d0);
3018: return r;
1.55 noro 3019: }
3020:
1.56 noro 3021: /* product of monomials */
3022: /* XXX block order is not handled correctly */
3023:
1.55 noro 3024: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3025: {
1.56 noro 3026: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3027: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3028: Q *ctab_q;
3029: Q q,q1;
1.61 noro 3030: UINT c0,c1,c;
1.55 noro 3031: NM *p;
3032: NM m,t;
3033:
3034: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3035: if ( !m0 || !m1 ) return;
3036: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3037: NEWNM(m); d = DL(m);
1.56 noro 3038: if ( mod ) {
3039: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3040: } else
3041: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3042: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3043: homo = n&1 ? 1 : 0;
3044: if ( homo ) {
3045: /* offset of h-degree */
3046: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3047: PUT_EXP(DL(m),n-1,h);
3048: TD(DL(m)) = h;
1.61 noro 3049: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3050: }
3051: tab[0] = m;
3052: NEWNM(m); d = DL(m);
1.57 noro 3053: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3054: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3055: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3056: /* xi^a*(Di^k*xi^l)*Di^b */
3057: a += l; b += k;
1.56 noro 3058: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3059: if ( !k || !l ) {
3060: for ( j = 0; j < curlen; j++ )
1.56 noro 3061: if ( t = tab[j] ) {
3062: dt = DL(t);
3063: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3064: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3065: }
3066: curlen *= k+1;
3067: continue;
3068: }
3069: min = MIN(k,l);
1.56 noro 3070: if ( mod ) {
1.61 noro 3071: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3072: mkwcm(k,l,mod,ctab);
3073: } else {
3074: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3075: mkwc(k,l,ctab_q);
3076: }
1.57 noro 3077: for ( j = min; j >= 0; j-- ) {
1.56 noro 3078: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3079: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3080: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3081: if ( homo ) {
3082: TD(d) = s;
1.56 noro 3083: PUT_EXP(d,n-1,s-h);
1.55 noro 3084: } else TD(d) = h;
1.61 noro 3085: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3086: if ( mod ) c = ctab[j];
3087: else q = ctab_q[j];
1.57 noro 3088: p = tab+curlen*j;
3089: if ( j == 0 ) {
3090: for ( u = 0; u < curlen; u++, p++ ) {
3091: if ( tab[u] ) {
3092: ndl_addto(DL(tab[u]),d);
3093: if ( mod ) {
3094: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3095: } else {
3096: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3097: }
3098: }
1.56 noro 3099: }
1.57 noro 3100: } else {
3101: for ( u = 0; u < curlen; u++, p++ ) {
3102: if ( tab[u] ) {
3103: NEWNM(t);
3104: ndl_add(DL(tab[u]),d,DL(t));
3105: if ( mod ) {
3106: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3107: } else
3108: mulq(CQ(tab[u]),q,&CQ(t));
3109: *p = t;
3110: }
1.55 noro 3111: }
3112: }
3113: }
3114: curlen *= k+1;
3115: }
3116: FREENM(m);
3117: }
3118:
1.63 noro 3119: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3120: {
3121: NM mr,mr0;
3122: NMV m;
3123: UINT *d,*dt,*dm;
3124: int c,n,td,i,c1,c2,len;
3125: Q q;
3126: ND r;
3127:
3128: if ( !p ) return 0;
3129: else {
3130: n = NV(p); m = BDY(p);
3131: d = DL(m0);
3132: len = LEN(p);
3133: mr0 = 0;
3134: td = TD(d);
3135: c = CM(m0);
3136: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3137: NEXTNM(mr0,mr);
3138: CM(mr) = 1;
3139: ndl_add(DL(m),d,DL(mr));
3140: }
3141: NEXT(mr) = 0;
3142: MKND(NV(p),mr0,len,r);
3143: SG(r) = SG(p) + TD(d);
3144: return r;
3145: }
3146: }
3147:
1.55 noro 3148: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3149: {
3150: NM mr,mr0;
3151: NMV m;
1.61 noro 3152: UINT *d,*dt,*dm;
1.9 noro 3153: int c,n,td,i,c1,c2,len;
1.16 noro 3154: Q q;
1.9 noro 3155: ND r;
3156:
1.34 noro 3157: if ( !p ) return 0;
1.55 noro 3158: else if ( do_weyl )
1.71 noro 3159: if ( mod == -1 )
3160: error("ndv_mul_nm : not implemented (weyl)");
3161: else
3162: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3163: else {
3164: n = NV(p); m = BDY(p);
1.34 noro 3165: d = DL(m0);
1.14 noro 3166: len = LEN(p);
1.9 noro 3167: mr0 = 0;
1.34 noro 3168: td = TD(d);
1.71 noro 3169: if ( mod == -1 ) {
3170: c = CM(m0);
3171: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3172: NEXTNM(mr0,mr);
3173: CM(mr) = _mulsf(CM(m),c);
3174: ndl_add(DL(m),d,DL(mr));
3175: }
3176: } else if ( mod ) {
1.16 noro 3177: c = CM(m0);
3178: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3179: NEXTNM(mr0,mr);
3180: c1 = CM(m);
1.19 noro 3181: DMAR(c1,c,0,mod,c2);
1.16 noro 3182: CM(mr) = c2;
3183: ndl_add(DL(m),d,DL(mr));
3184: }
3185: } else {
3186: q = CQ(m0);
3187: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3188: NEXTNM(mr0,mr);
3189: mulq(CQ(m),q,&CQ(mr));
3190: ndl_add(DL(m),d,DL(mr));
3191: }
1.4 noro 3192: }
1.9 noro 3193: NEXT(mr) = 0;
1.31 noro 3194: MKND(NV(p),mr0,len,r);
1.34 noro 3195: SG(r) = SG(p) + TD(d);
1.9 noro 3196: return r;
1.4 noro 3197: }
3198: }
3199:
1.43 noro 3200: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3201: {
1.13 noro 3202: NMV m,mr,mr0,t;
3203: int len,i,k;
1.11 noro 3204:
1.61 noro 3205: if ( !p ) return;
3206: m = BDY(p); len = LEN(p);
3207: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3208: m = (NMV)((char *)mr0+(len-1)*oadv);
3209: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3210: t = (NMV)ALLOCA(nmv_adv);
3211: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3212: CQ(t) = CQ(m);
3213: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3214: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3215: CQ(mr) = CQ(t);
3216: ndl_copy(DL(t),DL(mr));
3217: }
3218: BDY(p) = mr0;
3219: }
3220:
3221: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3222: {
3223: NMV m,mr,mr0;
3224: int len,i;
3225: NDV r;
1.11 noro 3226:
1.61 noro 3227: if ( !p ) return 0;
3228: m = BDY(p); len = LEN(p);
3229: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3230: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3231: ndl_zero(DL(mr));
3232: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3233: CQ(mr) = CQ(m);
1.11 noro 3234: }
1.61 noro 3235: MKNDV(NV(p),mr0,len,r);
3236: SG(r) = SG(p);
3237: return r;
1.11 noro 3238: }
3239:
1.61 noro 3240: /* duplicate p */
3241:
3242: NDV ndv_dup(int mod,NDV p)
1.3 noro 3243: {
3244: NDV d;
1.61 noro 3245: NMV t,m,m0;
1.3 noro 3246: int i,len;
3247:
1.34 noro 3248: if ( !p ) return 0;
1.31 noro 3249: len = LEN(p);
1.34 noro 3250: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3251: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3252: ndl_copy(DL(t),DL(m));
1.16 noro 3253: CQ(m) = CQ(t);
1.3 noro 3254: }
3255: MKNDV(NV(p),m0,len,d);
1.23 noro 3256: SG(d) = SG(p);
3257: return d;
3258: }
3259:
1.63 noro 3260: ND nd_dup(ND p)
3261: {
3262: ND d;
3263: NM t,m,m0;
3264:
3265: if ( !p ) return 0;
3266: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3267: NEXTNM(m0,m);
3268: ndl_copy(DL(t),DL(m));
3269: CQ(m) = CQ(t);
3270: }
3271: if ( m0 ) NEXT(m) = 0;
3272: MKND(NV(p),m0,LEN(p),d);
3273: SG(d) = SG(p);
3274: return d;
3275: }
3276:
1.61 noro 3277: /* XXX if p->len == 0 then it represents 0 */
3278:
3279: void ndv_mod(int mod,NDV p)
3280: {
3281: NMV t,d;
3282: int r;
3283: int i,len,dlen;
1.71 noro 3284: Obj gfs;
1.61 noro 3285:
3286: if ( !p ) return;
3287: len = LEN(p);
3288: dlen = 0;
1.71 noro 3289: if ( mod == -1 )
3290: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3291: simp_ff((Obj)CP(t),&gfs);
3292: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3293: CM(d) = r;
3294: ndl_copy(DL(t),DL(d));
3295: NMV_ADV(d);
3296: dlen++;
3297: }
1.71 noro 3298: else
3299: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3300: r = rem(NM(CQ(t)),mod);
3301: if ( r ) {
3302: if ( SGN(CQ(t)) < 0 )
3303: r = mod-r;
3304: CM(d) = r;
3305: ndl_copy(DL(t),DL(d));
3306: NMV_ADV(d);
3307: dlen++;
3308: }
3309: }
1.61 noro 3310: LEN(p) = dlen;
3311: }
3312:
3313: NDV ptondv(VL vl,VL dvl,P p)
3314: {
3315: ND nd;
3316:
3317: nd = ptond(vl,dvl,p);
3318: return ndtondv(0,nd);
3319: }
3320:
3321: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3322: {
1.61 noro 3323: int n,i,j,k,e;
3324: VL tvl;
3325: V v;
3326: DCP dc;
3327: DCP *w;
3328: ND r,s,t,u;
3329: P x;
3330: int c;
3331: UINT *d;
1.23 noro 3332: NM m,m0;
1.61 noro 3333:
3334: if ( !p )
3335: return 0;
3336: else if ( NUM(p) ) {
3337: NEWNM(m);
3338: ndl_zero(DL(m));
3339: CQ(m) = (Q)p;
3340: NEXT(m) = 0;
3341: MKND(nd_nvar,m,1,r);
3342: SG(r) = 0;
3343: return r;
3344: } else {
3345: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3346: w = (DCP *)ALLOCA(k*sizeof(DCP));
3347: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3348: for ( i = 0, tvl = dvl, v = VR(p);
3349: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3350: if ( !tvl ) {
3351: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3352: t = ptond(vl,dvl,COEF(w[j]));
3353: pwrp(vl,x,DEG(w[j]),&p);
3354: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3355: }
3356: return s;
3357: } else {
3358: NEWNM(m0); d = DL(m0);
3359: for ( j = k-1, s = 0; j >= 0; j-- ) {
3360: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3361: TD(d) = MUL_WEIGHT(e,i);
3362: if ( nd_blockmask) ndl_weight_mask(d);
3363: t = ptond(vl,dvl,COEF(w[j]));
3364: for ( m = BDY(t); m; m = NEXT(m) )
3365: ndl_addto(DL(m),d);
3366: SG(t) += TD(d);
3367: s = nd_add(0,s,t);
3368: }
3369: FREENM(m0);
3370: return s;
3371: }
3372: }
3373: }
3374:
3375: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3376: {
3377: VL tvl;
3378: int len,n,j,i,e;
3379: NMV m;
3380: Q q;
3381: P c;
3382: UINT *d;
3383: P s,r,u,t,w;
1.71 noro 3384: GFS gfs;
1.23 noro 3385:
1.34 noro 3386: if ( !p ) return 0;
1.61 noro 3387: else {
3388: len = LEN(p);
3389: n = NV(p);
3390: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3391: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3392: if ( mod == -1 ) {
3393: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3394: } else if ( mod ) {
1.61 noro 3395: STOQ(CM(m),q); c = (P)q;
3396: } else
3397: c = CP(m);
3398: d = DL(m);
3399: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3400: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3401: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3402: }
3403: addp(vl,s,t,&u); s = u;
3404: }
3405: return s;
1.23 noro 3406: }
1.3 noro 3407: }
3408:
1.61 noro 3409: NDV ndtondv(int mod,ND p)
1.11 noro 3410: {
3411: NDV d;
1.61 noro 3412: NMV m,m0;
3413: NM t;
3414: int i,len;
1.11 noro 3415:
1.34 noro 3416: if ( !p ) return 0;
1.61 noro 3417: len = LEN(p);
1.74 noro 3418: if ( mod )
3419: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
3420: else
3421: m0 = m = MALLOC(len*nmv_adv);
3422: ndv_alloc += nmv_adv*len;
1.61 noro 3423: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3424: ndl_copy(DL(t),DL(m));
3425: CQ(m) = CQ(t);
1.11 noro 3426: }
1.61 noro 3427: MKNDV(NV(p),m0,len,d);
1.14 noro 3428: SG(d) = SG(p);
1.11 noro 3429: return d;
3430: }
3431:
1.61 noro 3432: ND ndvtond(int mod,NDV p)
1.11 noro 3433: {
1.61 noro 3434: ND d;
3435: NM m,m0;
1.11 noro 3436: NMV t;
1.61 noro 3437: int i,len;
1.11 noro 3438:
1.34 noro 3439: if ( !p ) return 0;
1.11 noro 3440: m0 = 0;
1.61 noro 3441: len = p->len;
3442: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3443: NEXTNM(m0,m);
3444: ndl_copy(DL(t),DL(m));
3445: CQ(m) = CQ(t);
1.11 noro 3446: }
3447: NEXT(m) = 0;
1.61 noro 3448: MKND(NV(p),m0,len,d);
1.14 noro 3449: SG(d) = SG(p);
1.11 noro 3450: return d;
3451: }
3452:
1.3 noro 3453: void ndv_print(NDV p)
3454: {
3455: NMV m;
3456: int i,len;
3457:
1.34 noro 3458: if ( !p ) printf("0\n");
1.3 noro 3459: else {
1.14 noro 3460: len = LEN(p);
1.3 noro 3461: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3462: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3463: else printf("+%d*",CM(m));
1.16 noro 3464: ndl_print(DL(m));
3465: }
3466: printf("\n");
3467: }
3468: }
3469:
3470: void ndv_print_q(NDV p)
3471: {
3472: NMV m;
3473: int i,len;
3474:
1.34 noro 3475: if ( !p ) printf("0\n");
1.16 noro 3476: else {
3477: len = LEN(p);
3478: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3479: printf("+");
1.74 noro 3480: printexpr(CO,(Obj)CQ(m));
1.16 noro 3481: printf("*");
1.14 noro 3482: ndl_print(DL(m));
1.3 noro 3483: }
3484: printf("\n");
3485: }
1.25 noro 3486: }
3487:
1.61 noro 3488: NODE ndv_reducebase(NODE x)
1.27 noro 3489: {
3490: int len,i,j;
3491: NDV *w;
3492: NODE t,t0;
3493:
3494: len = length(x);
3495: w = (NDV *)ALLOCA(len*sizeof(NDV));
3496: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3497: for ( i = 0; i < len; i++ ) {
3498: for ( j = 0; j < i; j++ ) {
3499: if ( w[i] && w[j] )
3500: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3501: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3502: }
3503: }
3504: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3505: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3506: }
3507: NEXT(t) = 0; x = t0;
3508: return x;
1.11 noro 3509: }
1.32 noro 3510:
1.43 noro 3511: /* XXX incomplete */
3512:
1.32 noro 3513: void nd_init_ord(struct order_spec *ord)
3514: {
1.43 noro 3515: switch ( ord->id ) {
1.32 noro 3516: case 0:
1.43 noro 3517: switch ( ord->ord.simple ) {
3518: case 0:
3519: nd_dcomp = 1;
3520: nd_isrlex = 1;
3521: break;
3522: case 1:
3523: nd_dcomp = 1;
3524: nd_isrlex = 0;
3525: break;
3526: case 2:
3527: nd_dcomp = 0;
3528: nd_isrlex = 0;
1.45 noro 3529: ndl_compare_function = ndl_lex_compare;
1.58 noro 3530: break;
3531: case 11:
3532: /* XXX */
3533: nd_dcomp = 0;
3534: nd_isrlex = 1;
3535: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3536: break;
3537: default:
3538: error("nd_gr : unsupported order");
3539: }
1.32 noro 3540: break;
3541: case 1:
1.43 noro 3542: /* XXX */
3543: nd_dcomp = -1;
1.32 noro 3544: nd_isrlex = 0;
1.45 noro 3545: ndl_compare_function = ndl_block_compare;
1.34 noro 3546: break;
1.43 noro 3547: case 2:
3548: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3549: break;
3550: }
1.41 noro 3551: nd_ord = ord;
1.32 noro 3552: }
3553:
1.43 noro 3554: BlockMask nd_create_blockmask(struct order_spec *ord)
3555: {
3556: int n,i,j,s,l;
1.61 noro 3557: UINT *t;
1.43 noro 3558: BlockMask bm;
3559:
3560: if ( !ord->id )
3561: return 0;
3562: n = ord->ord.block.length;
3563: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3564: bm->n = n;
3565: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3566: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3567: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3568: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3569: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3570: l = bm->order_pair[i].length;
3571: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3572: }
3573: return bm;
1.57 noro 3574: }
3575:
3576: EPOS nd_create_epos(struct order_spec *ord)
3577: {
3578: int i,j,l,s,ord_l,ord_o;
3579: EPOS epos;
3580: struct order_pair *op;
3581:
3582: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3583: switch ( ord->id ) {
3584: case 0:
3585: if ( nd_isrlex ) {
3586: for ( i = 0; i < nd_nvar; i++ ) {
3587: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3588: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3589: }
3590: } else {
3591: for ( i = 0; i < nd_nvar; i++ ) {
3592: epos[i].i = nd_exporigin + i/nd_epw;
3593: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3594: }
3595: }
3596: break;
3597: case 1:
3598: /* block order */
3599: l = ord->ord.block.length;
3600: op = ord->ord.block.order_pair;
3601: for ( j = 0, s = 0; j < l; j++ ) {
3602: ord_o = op[j].order;
3603: ord_l = op[j].length;
3604: if ( !ord_o )
3605: for ( i = 0; i < ord_l; i++ ) {
3606: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3607: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3608: }
3609: else
3610: for ( i = 0; i < ord_l; i++ ) {
3611: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3612: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3613: }
3614: s += ord_l;
3615: }
3616: break;
3617: case 2:
3618: error("nd_create_epos : matrix order is not supported yet.");
3619: }
3620: return epos;
1.43 noro 3621: }
1.59 noro 3622:
3623: /* external interface */
3624:
3625: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3626: {
1.61 noro 3627: NODE t,in0,in;
1.59 noro 3628: ND nd,nf;
1.61 noro 3629: NDV ndv;
3630: VL vv,tv;
3631: int stat,nvar,max,e;
1.69 noro 3632: union oNDC dn;
1.59 noro 3633:
3634: pltovl(v,&vv);
1.74 noro 3635: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3636:
3637: /* get the degree bound */
3638: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3639: for ( tv = vv; tv; tv = NEXT(tv) ) {
3640: e = getdeg(tv->v,(P)BDY(t));
3641: max = MAX(e,max);
3642: }
3643: for ( tv = vv; tv; tv = NEXT(tv) ) {
3644: e = getdeg(tv->v,f);
3645: max = MAX(e,max);
3646: }
3647:
1.59 noro 3648: nd_init_ord(ord);
1.61 noro 3649: nd_setup_parameters(nvar,max);
3650:
3651: /* conversion to ndv */
3652: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3653: NEXTNODE(in0,in);
3654: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3655: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3656: }
3657: NEXTNODE(in0,in);
3658: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3659: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3660: NEXT(in) = 0;
3661:
3662: ndv_setup(m,0,in0);
1.59 noro 3663: nd_psn--;
3664: nd_scale=2;
3665: while ( 1 ) {
3666: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3667: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3668: if ( !stat ) {
3669: nd_psn++;
3670: nd_reconstruct(m,0,0);
3671: nd_psn--;
3672: } else
3673: break;
3674: }
1.61 noro 3675: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3676: }
3677:
3678: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3679: {
3680: NM m;
3681: UINT *t,*s;
3682: int i;
3683:
3684: for ( i = 0; i < n; i++ ) r[i] = 0;
3685: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3686: t = DL(m);
3687: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3688: r[i] = CM(m);
3689: }
3690: for ( i = 0; !r[i]; i++ );
3691: return i;
3692: }
3693:
1.74 noro 3694: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
3695: {
3696: NMV m;
3697: UINT *t,*s;
3698: int i,j,len;
3699:
3700: for ( i = 0; i < n; i++ ) r[i] = 0;
3701: m = BDY(d);
3702: len = LEN(d);
3703: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
3704: t = DL(m);
3705: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3706: r[i] = CM(m);
3707: }
3708: for ( i = 0; !r[i]; i++ );
3709: return i;
3710: }
3711:
1.63 noro 3712: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3713: {
3714: NM m;
3715: NMV mr;
3716: UINT *d,*t,*s;
3717: NDV p;
3718: int i,j,len;
3719:
3720: m = pair->mul;
3721: d = DL(m);
3722: p = nd_ps[pair->index];
3723: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3724: for ( i = 0; i < n; i++ ) r[i] = 0;
3725: len = LEN(p);
3726: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3727: ndl_add(d,DL(mr),t);
3728: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3729: r[i] = CM(mr);
3730: }
3731: for ( i = 0; !r[i]; i++ );
3732: return i;
3733: }
3734:
1.67 noro 3735: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3736: {
3737: NM m;
3738: NMV mr;
3739: UINT *d,*t,*s;
3740: NDV p;
1.67 noro 3741: unsigned char *ivc;
3742: unsigned short *ivs;
3743: UINT *v,*ivi;
3744: int i,j,len,prev,diff,cdiff;
3745: IndArray r;
1.64 noro 3746:
3747: m = pair->mul;
3748: d = DL(m);
3749: p = nd_ps[pair->index];
3750: len = LEN(p);
3751: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3752: r = (IndArray)MALLOC(sizeof(struct oIndArray));
3753: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3754: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3755: ndl_add(d,DL(mr),t);
3756: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3757: v[j] = i;
3758: }
3759: r->head = v[0];
3760: diff = 0;
3761: for ( i = 1; i < len; i++ ) {
3762: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3763: }
3764: if ( diff < 256 ) {
3765: r->width = 1;
3766: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3767: r->index.c = ivc;
3768: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3769: } else if ( diff < 65536 ) {
3770: r->width = 2;
3771: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3772: r->index.s = ivs;
3773: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3774: } else {
3775: r->width = 4;
3776: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3777: r->index.i = ivi;
3778: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3779: }
1.67 noro 3780: return r;
1.64 noro 3781: }
3782:
1.65 noro 3783:
1.76 ! noro 3784: int ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3785: {
1.67 noro 3786: int i,j,k,len,pos,prev;
1.66 noro 3787: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3788: IndArray ivect;
3789: unsigned char *ivc;
3790: unsigned short *ivs;
3791: unsigned int *ivi;
1.65 noro 3792: NDV redv;
1.67 noro 3793: NMV mr;
1.65 noro 3794: NODE rp;
1.76 ! noro 3795: int maxrs;
1.65 noro 3796:
1.76 ! noro 3797: maxrs = 0;
1.74 noro 3798: for ( i = 0; i < nred; i++ ) {
1.65 noro 3799: ivect = imat[i];
1.67 noro 3800: k = ivect->head; svect[k] %= m;
1.65 noro 3801: if ( c = svect[k] ) {
1.76 ! noro 3802: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3803: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3804: len = LEN(redv); mr = BDY(redv);
3805: svect[k] = 0; prev = k;
3806: switch ( ivect->width ) {
3807: case 1:
3808: ivc = ivect->index.c;
3809: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3810: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3811: prev = pos;
3812: DMA(c1,c,c2,up,lo);
3813: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3814: } else svect[pos] = lo;
3815: }
3816: break;
3817: case 2:
3818: ivs = ivect->index.s;
3819: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3820: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3821: prev = pos;
3822: DMA(c1,c,c2,up,lo);
3823: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3824: } else svect[pos] = lo;
3825: }
3826: break;
3827: case 4:
3828: ivi = ivect->index.i;
3829: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3830: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3831: prev = pos;
3832: DMA(c1,c,c2,up,lo);
3833: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3834: } else svect[pos] = lo;
3835: }
3836: break;
1.65 noro 3837: }
3838: }
3839: }
1.66 noro 3840: for ( i = 0; i < col; i++ )
3841: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.76 ! noro 3842: return maxrs;
1.65 noro 3843: }
3844:
1.76 ! noro 3845: int ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3846: {
3847: int i,j,k,len,pos,prev;
3848: UINT c,c1,c2,c3,up,lo,dmy;
3849: IndArray ivect;
3850: unsigned char *ivc;
3851: unsigned short *ivs;
3852: unsigned int *ivi;
3853: NDV redv;
3854: NMV mr;
3855: NODE rp;
1.76 ! noro 3856: int maxrs;
1.72 noro 3857:
1.76 ! noro 3858: maxrs = 0;
1.74 noro 3859: for ( i = 0; i < nred; i++ ) {
1.72 noro 3860: ivect = imat[i];
3861: k = ivect->head; svect[k] %= m;
3862: if ( c = svect[k] ) {
1.76 ! noro 3863: maxrs = MAX(maxrs,rp0[i]->sugar);
1.74 noro 3864: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3865: len = LEN(redv); mr = BDY(redv);
3866: svect[k] = 0; prev = k;
3867: switch ( ivect->width ) {
3868: case 1:
3869: ivc = ivect->index.c;
3870: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3871: pos = prev+ivc[j]; prev = pos;
3872: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3873: }
3874: break;
3875: case 2:
3876: ivs = ivect->index.s;
3877: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3878: pos = prev+ivs[j]; prev = pos;
3879: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3880: }
3881: break;
3882: case 4:
3883: ivi = ivect->index.i;
3884: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3885: pos = prev+ivi[j]; prev = pos;
3886: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3887: }
3888: break;
3889: }
3890: }
3891: }
1.76 ! noro 3892: return maxrs;
1.72 noro 3893: }
3894:
1.65 noro 3895: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3896: {
3897: int j,k,len;
3898: UINT *p;
3899: UINT c;
3900: NDV r;
3901: NMV mr0,mr;
3902:
3903: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3904: if ( !len ) return 0;
3905: else {
1.74 noro 3906: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
3907: ndv_alloc += nmv_adv*len;
1.65 noro 3908: mr = mr0;
3909: p = s0vect;
3910: for ( j = k = 0; j < col; j++, p += nd_wpd )
3911: if ( !rhead[j] ) {
3912: if ( c = vect[k++] ) {
3913: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3914: }
3915: }
3916: MKNDV(nd_nvar,mr0,len,r);
3917: return r;
3918: }
3919: }
3920:
1.74 noro 3921: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 3922: {
3923: ND_pairs t;
3924: NODE sp0,sp;
3925: int stat;
3926: ND spol;
3927:
3928: for ( t = l; t; t = NEXT(t) ) {
3929: stat = nd_sp(m,0,t,&spol);
3930: if ( !stat ) return 0;
3931: if ( spol ) {
3932: add_pbucket_symbolic(bucket,spol);
3933: }
3934: }
1.68 noro 3935: return 1;
1.65 noro 3936: }
3937:
3938: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
3939: {
3940: NODE rp0,rp;
3941: NM mul,head,s0,s;
1.76 ! noro 3942: int index,col,i,sugar;
1.65 noro 3943: RHist h;
3944: UINT *s0v,*p;
3945: NM_ind_pair pair;
3946: ND red;
3947:
3948: s0 = 0; rp0 = 0; col = 0;
3949: while ( 1 ) {
3950: head = remove_head_pbucket_symbolic(bucket);
3951: if ( !head ) break;
3952: if ( !s0 ) s0 = head;
3953: else NEXT(s) = head;
3954: s = head;
3955: index = ndl_find_reducer(DL(head));
3956: if ( index >= 0 ) {
3957: h = nd_psh[index];
3958: NEWNM(mul);
3959: ndl_sub(DL(head),DL(h),DL(mul));
3960: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
1.76 ! noro 3961: sugar = TD(DL(mul))+SG(nd_ps[index]);
! 3962: MKNM_ind_pair(pair,mul,index,sugar);
1.65 noro 3963: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
3964: add_pbucket_symbolic(bucket,nd_remove_head(red));
3965: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
3966: }
3967: col++;
3968: }
1.72 noro 3969: if ( rp0 ) NEXT(rp) = 0;
3970: NEXT(s) = 0;
1.65 noro 3971: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
3972: for ( i = 0, p = s0v, s = s0; i < col;
3973: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
3974: *s0vect = s0v;
3975: *r = rp0;
3976: return col;
3977: }
3978:
1.69 noro 3979: NODE nd_f4(int m)
3980: {
3981: int i,nh,stat,index;
3982: NODE r,g;
3983: ND_pairs d,l,t;
3984: ND spol,red;
3985: NDV nf,redv;
3986: NM s0,s;
1.74 noro 3987: NODE rp0,sp0,srp0,nflist;
1.69 noro 3988: int nsp,nred,col,rank,len,k,j,a;
3989: UINT c;
1.74 noro 3990: UINT **spmat;
1.69 noro 3991: UINT *s0vect,*svect,*p,*v;
3992: int *colstat;
3993: IndArray *imat;
3994: int *rhead;
3995: int spcol,sprow;
3996: int sugar;
3997: PGeoBucket bucket;
3998: struct oEGT eg0,eg1,eg_f4;
3999:
4000: if ( !m )
4001: error("nd_f4 : not implemented");
1.74 noro 4002: ndv_alloc = 0;
1.69 noro 4003: g = 0; d = 0;
4004: for ( i = 0; i < nd_psn; i++ ) {
4005: d = update_pairs(d,g,i);
4006: g = update_base(g,i);
4007: }
4008: while ( d ) {
4009: get_eg(&eg0);
4010: l = nd_minsugarp(d,&d);
4011: sugar = SG(l);
4012: bucket = create_pbucket();
1.74 noro 4013: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4014: if ( !stat ) {
4015: for ( t = l; NEXT(t); t = NEXT(t) );
4016: NEXT(t) = d; d = l;
4017: d = nd_reconstruct(m,0,d);
4018: continue;
4019: }
4020: if ( !sp0 ) continue;
4021: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4022: if ( !col ) {
4023: for ( t = l; NEXT(t); t = NEXT(t) );
4024: NEXT(t) = d; d = l;
4025: d = nd_reconstruct(m,0,d);
4026: continue;
4027: }
4028: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 noro 4029: if ( DP_Print )
4030: fprintf(asir_out,"sugar=%d,symb=%fsec,",
4031: sugar,eg_f4.exectime+eg_f4.gctime);
4032: if ( 1 )
4033: nflist = nd_f4_red(m,l,s0vect,col,rp0);
4034: else
4035: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4036: /* adding new bases */
1.74 noro 4037: for ( r = nflist; r; r = NEXT(r) ) {
4038: nf = (NDV)BDY(r);
1.69 noro 4039: ndv_removecont(m,nf);
4040: nh = ndv_newps(nf,0);
4041: d = update_pairs(d,g,nh);
4042: g = update_base(g,nh);
4043: }
4044: }
4045: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 noro 4046: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 4047: return g;
4048: }
1.74 noro 4049:
4050: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4051: {
1.67 noro 4052: IndArray *imat;
1.74 noro 4053: int nsp,nred,spcol,sprow,a;
1.65 noro 4054: int *rhead;
1.74 noro 4055: int i,j,k,l,rank;
4056: NODE rp,r0,r;
4057: ND_pairs sp;
4058: ND spol;
4059: int **spmat;
4060: UINT *svect,*v;
4061: int *colstat;
1.66 noro 4062: struct oEGT eg0,eg1,eg_f4;
1.74 noro 4063: NM_ind_pair *rvect;
1.76 ! noro 4064: int maxrs;
! 4065: int *spsugar;
1.63 noro 4066:
1.74 noro 4067: get_eg(&eg0);
4068: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4069: nred = length(rp0); spcol = col-nred;
4070: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
4071: rhead = (int *)ALLOCA(col*sizeof(int));
4072: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4073:
1.74 noro 4074: /* construction of index arrays */
4075: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
4076: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
4077: rvect[i] = (NM_ind_pair)BDY(rp);
4078: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
4079: rhead[imat[i]->head] = 1;
4080: }
4081:
4082: /* elimination (1st step) */
4083: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
4084: svect = (UINT *)ALLOCA(col*sizeof(UINT));
1.76 ! noro 4085: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4086: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
4087: nd_sp(m,0,sp,&spol);
1.75 noro 4088: if ( !spol ) continue;
1.74 noro 4089: nd_to_vect(m,s0vect,col,spol,svect);
1.76 ! noro 4090: if ( m == -1 )
! 4091: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
! 4092: else
! 4093: maxrs = ndv_reduce_vect(m,svect,col,imat,rvect,nred);
1.74 noro 4094: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4095: if ( i < col ) {
4096: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
4097: for ( j = k = 0; j < col; j++ )
4098: if ( !rhead[j] ) v[k++] = svect[j];
1.76 ! noro 4099: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4100: sprow++;
4101: }
1.76 ! noro 4102: nd_free(spol);
1.74 noro 4103: }
4104: /* free index arrays */
4105: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
4106:
4107: /* elimination (2nd step) */
4108: colstat = (int *)ALLOCA(spcol*sizeof(int));
4109: if ( m == -1 )
1.76 ! noro 4110: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4111: else
1.76 ! noro 4112: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4113: r0 = 0;
4114: for ( i = 0; i < rank; i++ ) {
4115: NEXTNODE(r0,r); BDY(r) =
4116: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
1.76 ! noro 4117: SG((NDV)BDY(r)) = spsugar[i];
1.74 noro 4118: GC_free(spmat[i]);
4119: }
4120: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
4121: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
4122: if ( DP_Print ) {
4123: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
4124: nsp,nred,sprow,spcol,rank);
4125: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4126: }
1.74 noro 4127: return r0;
4128: }
4129:
4130: FILE *nd_write,*nd_read;
4131:
4132: void nd_send_int(int a) {
4133: write_int(nd_write,&a);
4134: }
4135:
4136: void nd_send_intarray(int *p,int len) {
4137: write_intarray(nd_write,p,len);
4138: }
4139:
4140: int nd_recv_int() {
4141: int a;
4142:
4143: read_int(nd_read,&a);
4144: return a;
4145: }
4146:
4147: void nd_recv_intarray(int *p,int len) {
4148: read_intarray(nd_read,p,len);
4149: }
4150:
4151: void nd_send_ndv(NDV p) {
4152: int len,i;
4153: NMV m;
4154:
4155: if ( !p ) nd_send_int(0);
4156: else {
4157: len = LEN(p);
4158: nd_send_int(len);
4159: m = BDY(p);
4160: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4161: nd_send_int(CM(m));
4162: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4163: }
1.74 noro 4164: }
4165: }
4166:
4167: void nd_send_nd(ND p) {
4168: int len,i;
4169: NM m;
4170:
4171: if ( !p ) nd_send_int(0);
4172: else {
4173: len = LEN(p);
4174: nd_send_int(len);
4175: m = BDY(p);
4176: for ( i = 0; i < len; i++, m = NEXT(m) ) {
4177: nd_send_int(CM(m));
4178: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4179: }
1.74 noro 4180: }
4181: }
1.65 noro 4182:
1.74 noro 4183: NDV nd_recv_ndv()
4184: {
4185: int len,i;
4186: NMV m,m0;
4187: NDV r;
1.65 noro 4188:
1.74 noro 4189: len = nd_recv_int();
4190: if ( !len ) return 0;
4191: else {
4192: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
4193: ndv_alloc += len*nmv_adv;
4194: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
4195: CM(m) = nd_recv_int();
4196: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4197: }
1.74 noro 4198: MKNDV(nd_nvar,m0,len,r);
4199: return r;
4200: }
4201: }
1.65 noro 4202:
1.74 noro 4203: int ox_exec_f4_red(Q proc)
4204: {
4205: Obj obj;
4206: STRING fname;
4207: NODE arg;
4208: int s;
4209: extern int ox_need_conv,ox_file_io;
4210:
4211: MKSTR(fname,"nd_exec_f4_red");
4212: arg = mknode(2,proc,fname);
4213: Pox_cmo_rpc(arg,&obj);
4214: s = get_ox_server_id(QTOS(proc));
4215: nd_write = iofp[s].out;
4216: nd_read = iofp[s].in;
4217: ox_need_conv = ox_file_io = 0;
4218: return s;
4219: }
4220:
4221: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
4222: {
4223: int nsp,nred;
4224: int i,rank,s;
4225: NODE rp,r0,r;
4226: ND_pairs sp;
4227: NM_ind_pair pair;
4228: NMV nmv;
4229: NM nm;
4230: NDV nf;
4231: Obj proc,dmy;
4232:
4233: ox_launch_main(0,0,&proc);
4234: s = ox_exec_f4_red((Q)proc);
4235:
4236: nd_send_int(m);
4237: nd_send_int(nd_nvar);
4238: nd_send_int(nd_bpe);
4239: nd_send_int(nd_wpd);
4240: nd_send_int(nmv_adv);
4241:
4242: saveobj(nd_write,dp_current_spec.obj); fflush(nd_write);
4243:
4244: nd_send_int(nd_psn);
4245: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
4246:
4247: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
4248: nd_send_int(nsp);
4249: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
4250: nd_send_int(sp->i1); nd_send_int(sp->i2);
4251: }
4252:
4253: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
4254:
4255: nred = length(rp0); nd_send_int(nred);
4256: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
4257: pair = (NM_ind_pair)BDY(rp);
4258: nd_send_int(pair->index);
4259: nd_send_intarray(pair->mul->dl,nd_wpd);
4260: }
4261: fflush(nd_write);
4262: rank = nd_recv_int();
4263: fprintf(asir_out,"rank=%d\n",rank);
4264: r0 = 0;
4265: for ( i = 0; i < rank; i++ ) {
4266: nf = nd_recv_ndv();
4267: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
4268: }
4269: Pox_shutdown(mknode(1,proc),&dmy);
4270: return r0;
4271: }
4272:
4273: /* server side */
4274:
4275: void nd_exec_f4_red_dist()
4276: {
4277: int m,i,nsp,col,s0size,nred,spcol,j,k;
4278: NM_ind_pair *rp0;
4279: NDV nf;
4280: UINT *s0vect;
4281: IndArray *imat;
4282: int *rhead;
4283: int **spmat;
4284: UINT *svect,*v;
4285: ND_pairs *sp0;
4286: int *colstat;
4287: int a,sprow,rank;
4288: struct order_spec ord;
4289: Obj ordspec;
4290: ND spol;
1.76 ! noro 4291: int maxrs;
! 4292: int *spsugar;
1.74 noro 4293:
4294: nd_read = iofp[0].in;
4295: nd_write = iofp[0].out;
4296: m = nd_recv_int();
4297: nd_nvar = nd_recv_int();
4298: nd_bpe = nd_recv_int();
4299: nd_wpd = nd_recv_int();
4300: nmv_adv = nd_recv_int();
4301:
4302: loadobj(nd_read,&ordspec);
4303: create_order_spec(ordspec,&ord);
4304: nd_init_ord(&ord);
4305: nd_setup_parameters(nd_nvar,0);
4306:
4307: nd_psn = nd_recv_int();
4308: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
4309: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
4310: for ( i = 0; i < nd_psn; i++ ) {
4311: nd_ps[i] = nd_recv_ndv();
4312: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
4313: }
4314:
4315: nsp = nd_recv_int();
4316: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
4317: for ( i = 0; i < nsp; i++ ) {
4318: NEWND_pairs(sp0[i]);
4319: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
4320: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
4321: }
4322:
4323: col = nd_recv_int();
4324: s0size = col*nd_wpd;
4325: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
4326: nd_recv_intarray(s0vect,s0size);
4327:
4328: nred = nd_recv_int();
4329: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
4330: for ( i = 0; i < nred; i++ ) {
4331: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
4332: rp0[i]->index = nd_recv_int();
4333: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
4334: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
4335: }
4336:
4337: spcol = col-nred;
4338: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
4339: rhead = (int *)MALLOC(col*sizeof(int));
4340: for ( i = 0; i < col; i++ ) rhead[i] = 0;
4341:
4342: /* construction of index arrays */
4343: for ( i = 0; i < nred; i++ ) {
4344: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
4345: rhead[imat[i]->head] = 1;
4346: }
4347:
4348: /* elimination (1st step) */
4349: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
4350: svect = (UINT *)MALLOC(col*sizeof(UINT));
1.76 ! noro 4351: spsugar = (int *)ALLOCA(nsp*sizeof(UINT));
1.74 noro 4352: for ( a = sprow = 0; a < nsp; a++ ) {
4353: nd_sp(m,0,sp0[a],&spol);
1.75 noro 4354: if ( !spol ) continue;
1.74 noro 4355: nd_to_vect(m,s0vect,col,spol,svect);
1.76 ! noro 4356: if ( m == -1 )
! 4357: maxrs = ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
! 4358: else
! 4359: maxrs = ndv_reduce_vect(m,svect,col,imat,rp0,nred);
1.74 noro 4360: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
4361: if ( i < col ) {
4362: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
4363: for ( j = k = 0; j < col; j++ )
4364: if ( !rhead[j] ) v[k++] = svect[j];
1.76 ! noro 4365: spsugar[sprow] = MAX(maxrs,SG(spol));
1.74 noro 4366: sprow++;
4367: }
1.76 ! noro 4368: nd_free(spol);
1.74 noro 4369: }
4370: /* elimination (2nd step) */
4371: colstat = (int *)ALLOCA(spcol*sizeof(int));
4372: if ( m == -1 )
1.76 ! noro 4373: rank = nd_gauss_elim_sf(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4374: else
1.76 ! noro 4375: rank = nd_gauss_elim_mod(spmat,spsugar,sprow,spcol,m,colstat);
1.74 noro 4376: nd_send_int(rank);
4377: for ( i = 0; i < rank; i++ ) {
4378: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
4379: nd_send_ndv(nf);
1.63 noro 4380: }
1.74 noro 4381: fflush(nd_write);
1.76 ! noro 4382: }
! 4383:
! 4384: int nd_gauss_elim_mod(int **mat0,int *sugar,int row,int col,int md,int *colstat)
! 4385: {
! 4386: int i,j,k,l,inv,a,rank,s;
! 4387: unsigned int *t,*pivot,*pk;
! 4388: unsigned int **mat;
! 4389:
! 4390: mat = (unsigned int **)mat0;
! 4391: for ( rank = 0, j = 0; j < col; j++ ) {
! 4392: for ( i = rank; i < row; i++ )
! 4393: mat[i][j] %= md;
! 4394: for ( i = rank; i < row; i++ )
! 4395: if ( mat[i][j] )
! 4396: break;
! 4397: if ( i == row ) {
! 4398: colstat[j] = 0;
! 4399: continue;
! 4400: } else
! 4401: colstat[j] = 1;
! 4402: if ( i != rank ) {
! 4403: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
! 4404: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
! 4405: }
! 4406: pivot = mat[rank];
! 4407: s = sugar[rank];
! 4408: inv = invm(pivot[j],md);
! 4409: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
! 4410: if ( *pk ) {
! 4411: if ( *pk >= (unsigned int)md )
! 4412: *pk %= md;
! 4413: DMAR(*pk,inv,0,md,*pk)
! 4414: }
! 4415: for ( i = rank+1; i < row; i++ ) {
! 4416: t = mat[i];
! 4417: if ( a = t[j] ) {
! 4418: sugar[i] = MAX(sugar[i],s);
! 4419: red_by_vect(md,t+j,pivot+j,md-a,col-j);
! 4420: }
! 4421: }
! 4422: rank++;
! 4423: }
! 4424: for ( j = col-1, l = rank-1; j >= 0; j-- )
! 4425: if ( colstat[j] ) {
! 4426: pivot = mat[l];
! 4427: s = sugar[l];
! 4428: for ( i = 0; i < l; i++ ) {
! 4429: t = mat[i];
! 4430: t[j] %= md;
! 4431: if ( a = t[j] ) {
! 4432: sugar[i] = MAX(sugar[i],s);
! 4433: red_by_vect(md,t+j,pivot+j,md-a,col-j);
! 4434: }
! 4435: }
! 4436: l--;
! 4437: }
! 4438: for ( j = 0, l = 0; l < rank; j++ )
! 4439: if ( colstat[j] ) {
! 4440: t = mat[l];
! 4441: for ( k = j; k < col; k++ )
! 4442: if ( t[k] >= (unsigned int)md )
! 4443: t[k] %= md;
! 4444: l++;
! 4445: }
! 4446: return rank;
! 4447: }
! 4448:
! 4449: int nd_gauss_elim_sf(int **mat0,int *sugar,int row,int col,int md,int *colstat)
! 4450: {
! 4451: int i,j,k,l,inv,a,rank,s;
! 4452: unsigned int *t,*pivot,*pk;
! 4453: unsigned int **mat;
! 4454:
! 4455: mat = (unsigned int **)mat0;
! 4456: for ( rank = 0, j = 0; j < col; j++ ) {
! 4457: for ( i = rank; i < row; i++ )
! 4458: if ( mat[i][j] )
! 4459: break;
! 4460: if ( i == row ) {
! 4461: colstat[j] = 0;
! 4462: continue;
! 4463: } else
! 4464: colstat[j] = 1;
! 4465: if ( i != rank ) {
! 4466: t = mat[i]; mat[i] = mat[rank]; mat[rank] = t;
! 4467: s = sugar[i]; sugar[i] = sugar[rank]; sugar[rank] = s;
! 4468: }
! 4469: pivot = mat[rank];
! 4470: s = sugar[rank];
! 4471: inv = _invsf(pivot[j]);
! 4472: for ( k = j, pk = pivot+k; k < col; k++, pk++ )
! 4473: if ( *pk )
! 4474: *pk = _mulsf(*pk,inv);
! 4475: for ( i = rank+1; i < row; i++ ) {
! 4476: t = mat[i];
! 4477: if ( a = t[j] ) {
! 4478: sugar[i] = MAX(sugar[i],s);
! 4479: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
! 4480: }
! 4481: }
! 4482: rank++;
! 4483: }
! 4484: for ( j = col-1, l = rank-1; j >= 0; j-- )
! 4485: if ( colstat[j] ) {
! 4486: pivot = mat[l];
! 4487: s = sugar[l];
! 4488: for ( i = 0; i < l; i++ ) {
! 4489: t = mat[i];
! 4490: if ( a = t[j] ) {
! 4491: sugar[i] = MAX(sugar[i],s);
! 4492: red_by_vect_sf(md,t+j,pivot+j,_chsgnsf(a),col-j);
! 4493: }
! 4494: }
! 4495: l--;
! 4496: }
! 4497: return rank;
1.59 noro 4498: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>
![[BACK]](/icons/cvsweb/back.gif){kind=icon}
Return to nd.c CVS log ![[TXT]](/icons/cvsweb/text.gif){kind=icon}
![[DIR]](/icons/cvsweb/dir.gif){kind=icon}
Up to [local] / OpenXM_contrib2 / asir2000 / engine