Annotation of OpenXM_contrib2/asir2000/engine/nd.c, Revision 1.74
1.74 ! noro 1: /* $OpenXM: OpenXM_contrib2/asir2000/engine/nd.c,v 1.73 2003/09/17 08:37:49 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;
103: int index;
104: } *NM_ind_pair;
105:
1.67 noro 106: typedef struct oIndArray
107: {
108: char width;
109: int head;
110: union {
111: unsigned char *c;
112: unsigned short *s;
113: unsigned int *i;
114: } index;
115: } *IndArray;
1.63 noro 116:
1.61 noro 117: int (*ndl_compare_function)(UINT *a1,UINT *a2);
1.32 noro 118:
1.74 ! 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);}
1.63 noro 216: #define MKNM_ind_pair(p,m,i) (NEWNM_ind_pair(p),(p)->mul=(m),(p)->index=(i))
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:
371: void nd_free_private_storage()
372: {
373: _nm_free_list = 0;
1.5 noro 374: _ndp_free_list = 0;
1.71 noro 375: #if 0
1.1 noro 376: GC_gcollect();
1.71 noro 377: #endif
1.1 noro 378: }
379:
380: void _NM_alloc()
381: {
382: NM p;
383: int i;
384:
1.11 noro 385: for ( i = 0; i < 1024; i++ ) {
1.61 noro 386: p = (NM)GC_malloc(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 387: p->next = _nm_free_list; _nm_free_list = p;
388: }
389: }
390:
391: void _ND_alloc()
392: {
393: ND p;
394: int i;
395:
396: for ( i = 0; i < 1024; i++ ) {
397: p = (ND)GC_malloc(sizeof(struct oND));
398: p->body = (NM)_nd_free_list; _nd_free_list = p;
399: }
400: }
401:
402: void _NDP_alloc()
403: {
404: ND_pairs p;
405: int i;
406:
1.11 noro 407: for ( i = 0; i < 1024; i++ ) {
1.1 noro 408: p = (ND_pairs)GC_malloc(sizeof(struct oND_pairs)
1.61 noro 409: +(nd_wpd-1)*sizeof(UINT));
1.1 noro 410: p->next = _ndp_free_list; _ndp_free_list = p;
411: }
412: }
413:
1.30 noro 414: INLINE int nd_length(ND p)
1.1 noro 415: {
416: NM m;
417: int i;
418:
419: if ( !p )
420: return 0;
421: else {
422: for ( i = 0, m = BDY(p); m; m = NEXT(m), i++ );
423: return i;
424: }
425: }
426:
1.61 noro 427: INLINE int ndl_reducible(UINT *d1,UINT *d2)
1.1 noro 428: {
1.61 noro 429: UINT u1,u2;
1.1 noro 430: int i,j;
431:
1.34 noro 432: if ( TD(d1) < TD(d2) ) return 0;
1.65 noro 433: #if USE_UNROLL
1.1 noro 434: switch ( nd_bpe ) {
1.62 noro 435: case 3:
436: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
437: u1 = d1[i]; u2 = d2[i];
438: if ( (u1&0x38000000) < (u2&0x38000000) ) return 0;
439: if ( (u1& 0x7000000) < (u2& 0x7000000) ) return 0;
440: if ( (u1& 0xe00000) < (u2& 0xe00000) ) return 0;
441: if ( (u1& 0x1c0000) < (u2& 0x1c0000) ) return 0;
442: if ( (u1& 0x38000) < (u2& 0x38000) ) return 0;
443: if ( (u1& 0x7000) < (u2& 0x7000) ) return 0;
444: if ( (u1& 0xe00) < (u2& 0xe00) ) return 0;
445: if ( (u1& 0x1c0) < (u2& 0x1c0) ) return 0;
446: if ( (u1& 0x38) < (u2& 0x38) ) return 0;
447: if ( (u1& 0x7) < (u2& 0x7) ) return 0;
448: }
449: return 1;
450: break;
1.1 noro 451: case 4:
1.41 noro 452: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 453: u1 = d1[i]; u2 = d2[i];
454: if ( (u1&0xf0000000) < (u2&0xf0000000) ) return 0;
1.62 noro 455: if ( (u1& 0xf000000) < (u2& 0xf000000) ) return 0;
456: if ( (u1& 0xf00000) < (u2& 0xf00000) ) return 0;
457: if ( (u1& 0xf0000) < (u2& 0xf0000) ) return 0;
458: if ( (u1& 0xf000) < (u2& 0xf000) ) return 0;
459: if ( (u1& 0xf00) < (u2& 0xf00) ) return 0;
460: if ( (u1& 0xf0) < (u2& 0xf0) ) return 0;
461: if ( (u1& 0xf) < (u2& 0xf) ) return 0;
1.1 noro 462: }
463: return 1;
464: break;
465: case 6:
1.41 noro 466: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 467: u1 = d1[i]; u2 = d2[i];
468: if ( (u1&0x3f000000) < (u2&0x3f000000) ) return 0;
1.62 noro 469: if ( (u1& 0xfc0000) < (u2& 0xfc0000) ) return 0;
470: if ( (u1& 0x3f000) < (u2& 0x3f000) ) return 0;
471: if ( (u1& 0xfc0) < (u2& 0xfc0) ) return 0;
472: if ( (u1& 0x3f) < (u2& 0x3f) ) return 0;
1.1 noro 473: }
474: return 1;
475: break;
476: case 8:
1.41 noro 477: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 478: u1 = d1[i]; u2 = d2[i];
479: if ( (u1&0xff000000) < (u2&0xff000000) ) return 0;
1.62 noro 480: if ( (u1& 0xff0000) < (u2& 0xff0000) ) return 0;
481: if ( (u1& 0xff00) < (u2& 0xff00) ) return 0;
482: if ( (u1& 0xff) < (u2& 0xff) ) return 0;
1.1 noro 483: }
484: return 1;
485: break;
486: case 16:
1.41 noro 487: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 488: u1 = d1[i]; u2 = d2[i];
489: if ( (u1&0xffff0000) < (u2&0xffff0000) ) return 0;
1.62 noro 490: if ( (u1& 0xffff) < (u2& 0xffff) ) return 0;
1.1 noro 491: }
492: return 1;
493: break;
494: case 32:
1.41 noro 495: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 496: if ( d1[i] < d2[i] ) return 0;
497: return 1;
498: break;
499: default:
1.41 noro 500: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 501: u1 = d1[i]; u2 = d2[i];
502: for ( j = 0; j < nd_epw; j++ )
503: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
504: }
505: return 1;
506: }
1.65 noro 507: #else
508: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
509: u1 = d1[i]; u2 = d2[i];
510: for ( j = 0; j < nd_epw; j++ )
511: if ( (u1&nd_mask[j]) < (u2&nd_mask[j]) ) return 0;
512: }
513: return 1;
514: #endif
1.1 noro 515: }
516:
1.61 noro 517: /*
518: * If the current order is a block order,
519: * then the last block is length 1 and contains
520: * the homo variable. Otherwise, the original
521: * order is either 0 or 2.
522: */
523:
524: void ndl_homogenize(UINT *d,UINT *r,int obpe,EPOS oepos,int weight)
1.23 noro 525: {
1.61 noro 526: int w,i,e,n,omask0;
527:
528: omask0 = (1<<obpe)-1;
529: n = nd_nvar-1;
530: ndl_zero(r);
531: for ( i = 0; i < n; i++ ) {
532: e = GET_EXP_OLD(d,i);
533: PUT_EXP(r,i,e);
534: }
535: w = TD(d);
536: PUT_EXP(r,nd_nvar-1,weight-w);
537: TD(r) = weight;
538: if ( nd_blockmask ) ndl_weight_mask(r);
539: }
540:
541: void ndl_dehomogenize(UINT *d)
542: {
543: UINT mask;
544: UINT h;
1.31 noro 545: int i,bits;
1.23 noro 546:
1.44 noro 547: if ( nd_blockmask ) {
548: h = GET_EXP(d,nd_nvar-1);
1.45 noro 549: XOR_EXP(d,nd_nvar-1,h);
1.44 noro 550: TD(d) -= h;
551: d[nd_exporigin-1] -= h;
552: } else {
553: if ( nd_isrlex ) {
554: if ( nd_bpe == 32 ) {
555: h = d[nd_exporigin];
556: for ( i = nd_exporigin+1; i < nd_wpd; i++ )
557: d[i-1] = d[i];
558: d[i-1] = 0;
559: TD(d) -= h;
560: } else {
561: bits = nd_epw*nd_bpe;
562: mask = bits==32?0xffffffff:((1<<(nd_epw*nd_bpe))-1);
563: h = (d[nd_exporigin]>>((nd_epw-1)*nd_bpe))&nd_mask0;
564: for ( i = nd_exporigin; i < nd_wpd; i++ )
565: d[i] = ((d[i]<<nd_bpe)&mask)
566: |(i+1<nd_wpd?((d[i+1]>>((nd_epw-1)*nd_bpe))&nd_mask0):0);
567: TD(d) -= h;
568: }
1.45 noro 569: } else {
570: h = GET_EXP(d,nd_nvar-1);
571: XOR_EXP(d,nd_nvar-1,h);
572: TD(d) -= h;
573: }
1.44 noro 574: }
1.23 noro 575: }
576:
1.61 noro 577: void ndl_lcm(UINT *d1,unsigned *d2,UINT *d)
1.1 noro 578: {
1.61 noro 579: UINT t1,t2,u,u1,u2;
1.43 noro 580: int i,j,l;
1.1 noro 581:
1.65 noro 582: #if USE_UNROLL
1.1 noro 583: switch ( nd_bpe ) {
1.62 noro 584: case 3:
585: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
586: u1 = d1[i]; u2 = d2[i];
587: t1 = (u1&0x38000000); t2 = (u2&0x38000000); u = t1>t2?t1:t2;
588: t1 = (u1& 0x7000000); t2 = (u2& 0x7000000); u |= t1>t2?t1:t2;
589: t1 = (u1& 0xe00000); t2 = (u2& 0xe00000); u |= t1>t2?t1:t2;
590: t1 = (u1& 0x1c0000); t2 = (u2& 0x1c0000); u |= t1>t2?t1:t2;
591: t1 = (u1& 0x38000); t2 = (u2& 0x38000); u |= t1>t2?t1:t2;
592: t1 = (u1& 0x7000); t2 = (u2& 0x7000); u |= t1>t2?t1:t2;
593: t1 = (u1& 0xe00); t2 = (u2& 0xe00); u |= t1>t2?t1:t2;
594: t1 = (u1& 0x1c0); t2 = (u2& 0x1c0); u |= t1>t2?t1:t2;
595: t1 = (u1& 0x38); t2 = (u2& 0x38); u |= t1>t2?t1:t2;
596: t1 = (u1& 0x7); t2 = (u2& 0x7); u |= t1>t2?t1:t2;
597: d[i] = u;
598: }
599: break;
1.1 noro 600: case 4:
1.41 noro 601: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 602: u1 = d1[i]; u2 = d2[i];
603: t1 = (u1&0xf0000000); t2 = (u2&0xf0000000); u = t1>t2?t1:t2;
1.62 noro 604: t1 = (u1& 0xf000000); t2 = (u2& 0xf000000); u |= t1>t2?t1:t2;
605: t1 = (u1& 0xf00000); t2 = (u2& 0xf00000); u |= t1>t2?t1:t2;
606: t1 = (u1& 0xf0000); t2 = (u2& 0xf0000); u |= t1>t2?t1:t2;
607: t1 = (u1& 0xf000); t2 = (u2& 0xf000); u |= t1>t2?t1:t2;
608: t1 = (u1& 0xf00); t2 = (u2& 0xf00); u |= t1>t2?t1:t2;
609: t1 = (u1& 0xf0); t2 = (u2& 0xf0); u |= t1>t2?t1:t2;
610: t1 = (u1& 0xf); t2 = (u2& 0xf); u |= t1>t2?t1:t2;
1.1 noro 611: d[i] = u;
612: }
613: break;
614: case 6:
1.41 noro 615: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 616: u1 = d1[i]; u2 = d2[i];
617: t1 = (u1&0x3f000000); t2 = (u2&0x3f000000); u = t1>t2?t1:t2;
1.62 noro 618: t1 = (u1& 0xfc0000); t2 = (u2& 0xfc0000); u |= t1>t2?t1:t2;
619: t1 = (u1& 0x3f000); t2 = (u2& 0x3f000); u |= t1>t2?t1:t2;
620: t1 = (u1& 0xfc0); t2 = (u2& 0xfc0); u |= t1>t2?t1:t2;
621: t1 = (u1& 0x3f); t2 = (u2& 0x3f); u |= t1>t2?t1:t2;
1.1 noro 622: d[i] = u;
623: }
624: break;
625: case 8:
1.41 noro 626: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 627: u1 = d1[i]; u2 = d2[i];
628: t1 = (u1&0xff000000); t2 = (u2&0xff000000); u = t1>t2?t1:t2;
1.62 noro 629: t1 = (u1& 0xff0000); t2 = (u2& 0xff0000); u |= t1>t2?t1:t2;
630: t1 = (u1& 0xff00); t2 = (u2& 0xff00); u |= t1>t2?t1:t2;
631: t1 = (u1& 0xff); t2 = (u2& 0xff); u |= t1>t2?t1:t2;
1.1 noro 632: d[i] = u;
633: }
634: break;
635: case 16:
1.41 noro 636: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 637: u1 = d1[i]; u2 = d2[i];
638: t1 = (u1&0xffff0000); t2 = (u2&0xffff0000); u = t1>t2?t1:t2;
1.62 noro 639: t1 = (u1& 0xffff); t2 = (u2& 0xffff); u |= t1>t2?t1:t2;
1.1 noro 640: d[i] = u;
641: }
642: break;
643: case 32:
1.41 noro 644: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 645: u1 = d1[i]; u2 = d2[i];
646: d[i] = u1>u2?u1:u2;
647: }
648: break;
649: default:
1.41 noro 650: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 651: u1 = d1[i]; u2 = d2[i];
652: for ( j = 0, u = 0; j < nd_epw; j++ ) {
653: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
654: }
655: d[i] = u;
656: }
657: break;
658: }
1.65 noro 659: #else
660: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
661: u1 = d1[i]; u2 = d2[i];
662: for ( j = 0, u = 0; j < nd_epw; j++ ) {
663: t1 = (u1&nd_mask[j]); t2 = (u2&nd_mask[j]); u |= t1>t2?t1:t2;
664: }
665: d[i] = u;
666: }
667: #endif
1.39 noro 668: TD(d) = ndl_weight(d);
1.61 noro 669: if ( nd_blockmask ) ndl_weight_mask(d);
1.57 noro 670: }
671:
1.61 noro 672: int ndl_weight(UINT *d)
1.1 noro 673: {
1.61 noro 674: UINT t,u;
1.1 noro 675: int i,j;
676:
1.60 noro 677: if ( current_dl_weight_vector )
678: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
679: u = GET_EXP(d,i);
680: t += MUL_WEIGHT(u,i);
681: }
682: else
683: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
684: u = d[i];
685: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
686: t += (u&nd_mask0);
687: }
1.1 noro 688: return t;
689: }
690:
1.61 noro 691: void ndl_weight_mask(UINT *d)
1.43 noro 692: {
1.61 noro 693: UINT t,u;
694: UINT *mask;
695: int i,j,k,l;
1.43 noro 696:
1.61 noro 697: l = nd_blockmask->n;
698: for ( k = 0; k < l; k++ ) {
699: mask = nd_blockmask->mask[k];
700: if ( current_dl_weight_vector )
701: for ( i = 0, t = 0; i < nd_nvar; i++ ) {
702: u = GET_EXP_MASK(d,i,mask);
703: t += MUL_WEIGHT(u,i);
704: }
705: else
706: for ( t = 0, i = nd_exporigin; i < nd_wpd; i++ ) {
707: u = d[i]&mask[i];
708: for ( j = 0; j < nd_epw; j++, u>>=nd_bpe )
709: t += (u&nd_mask0);
710: }
711: d[k+1] = t;
712: }
1.43 noro 713: }
714:
1.61 noro 715: int ndl_lex_compare(UINT *d1,UINT *d2)
1.1 noro 716: {
717: int i;
718:
1.41 noro 719: d1 += nd_exporigin;
720: d2 += nd_exporigin;
721: for ( i = nd_exporigin; i < nd_wpd; i++, d1++, d2++ )
1.1 noro 722: if ( *d1 > *d2 )
1.32 noro 723: return nd_isrlex ? -1 : 1;
1.1 noro 724: else if ( *d1 < *d2 )
1.32 noro 725: return nd_isrlex ? 1 : -1;
1.1 noro 726: return 0;
727: }
728:
1.61 noro 729: int ndl_block_compare(UINT *d1,UINT *d2)
1.43 noro 730: {
731: int i,l,j,ord_o,ord_l;
732: struct order_pair *op;
1.61 noro 733: UINT t1,t2,m;
734: UINT *mask;
1.43 noro 735:
736: l = nd_blockmask->n;
737: op = nd_blockmask->order_pair;
738: for ( j = 0; j < l; j++ ) {
739: mask = nd_blockmask->mask[j];
740: ord_o = op[j].order;
741: if ( ord_o < 2 )
1.44 noro 742: if ( (t1=d1[j+1]) > (t2=d2[j+1]) ) return 1;
743: else if ( t1 < t2 ) return -1;
1.43 noro 744: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.44 noro 745: m = mask[i];
746: t1 = d1[i]&m;
747: t2 = d2[i]&m;
1.43 noro 748: if ( t1 > t2 )
749: return !ord_o ? -1 : 1;
750: else if ( t1 < t2 )
751: return !ord_o ? 1 : -1;
752: }
753: }
754: return 0;
755: }
756:
1.58 noro 757: /* TDH -> WW -> TD-> RL */
758:
1.61 noro 759: int ndl_ww_lex_compare(UINT *d1,UINT *d2)
1.58 noro 760: {
761: int i,m,e1,e2;
762:
763: if ( TD(d1) > TD(d2) ) return 1;
764: else if ( TD(d1) < TD(d2) ) return -1;
765: m = nd_nvar>>1;
766: for ( i = 0, e1 = e2 = 0; i < m; i++ ) {
767: e1 += current_weyl_weight_vector[i]*(GET_EXP(d1,m+i)-GET_EXP(d1,i));
768: e2 += current_weyl_weight_vector[i]*(GET_EXP(d2,m+i)-GET_EXP(d2,i));
769: }
770: if ( e1 > e2 ) return 1;
771: else if ( e1 < e2 ) return -1;
772: return ndl_lex_compare(d1,d2);
773: }
774:
1.61 noro 775: INLINE int ndl_equal(UINT *d1,UINT *d2)
1.1 noro 776: {
777: int i;
778:
1.41 noro 779: for ( i = 0; i < nd_wpd; i++ )
1.34 noro 780: if ( *d1++ != *d2++ )
1.1 noro 781: return 0;
782: return 1;
783: }
784:
1.61 noro 785: INLINE void ndl_copy(UINT *d1,UINT *d2)
1.6 noro 786: {
787: int i;
788:
789: switch ( nd_wpd ) {
1.41 noro 790: case 2:
1.34 noro 791: TD(d2) = TD(d1);
792: d2[1] = d1[1];
1.6 noro 793: break;
1.41 noro 794: case 3:
1.34 noro 795: TD(d2) = TD(d1);
1.6 noro 796: d2[1] = d1[1];
1.34 noro 797: d2[2] = d1[2];
1.6 noro 798: break;
799: default:
1.41 noro 800: for ( i = 0; i < nd_wpd; i++ )
1.6 noro 801: d2[i] = d1[i];
802: break;
803: }
804: }
805:
1.61 noro 806: INLINE void ndl_zero(UINT *d)
807: {
808: int i;
809: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
810: }
811:
812: INLINE void ndl_add(UINT *d1,UINT *d2,UINT *d)
1.1 noro 813: {
814: int i;
815:
1.43 noro 816: #if 1
1.6 noro 817: switch ( nd_wpd ) {
1.41 noro 818: case 2:
819: TD(d) = TD(d1)+TD(d2);
1.34 noro 820: d[1] = d1[1]+d2[1];
1.6 noro 821: break;
1.41 noro 822: case 3:
823: TD(d) = TD(d1)+TD(d2);
1.6 noro 824: d[1] = d1[1]+d2[1];
1.34 noro 825: d[2] = d1[2]+d2[2];
1.6 noro 826: break;
827: default:
1.43 noro 828: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
1.6 noro 829: break;
830: }
1.43 noro 831: #else
832: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]+d2[i];
833: #endif
1.6 noro 834: }
835:
1.55 noro 836: /* d1 += d2 */
1.61 noro 837: INLINE void ndl_addto(UINT *d1,UINT *d2)
1.55 noro 838: {
839: int i;
840:
841: #if 1
842: switch ( nd_wpd ) {
843: case 2:
844: TD(d1) += TD(d2);
845: d1[1] += d2[1];
846: break;
847: case 3:
848: TD(d1) += TD(d2);
849: d1[1] += d2[1];
850: d1[2] += d2[2];
851: break;
852: default:
853: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
854: break;
855: }
856: #else
857: for ( i = 0; i < nd_wpd; i++ ) d1[i] += d2[i];
858: #endif
859: }
860:
1.61 noro 861: INLINE void ndl_sub(UINT *d1,UINT *d2,UINT *d)
1.6 noro 862: {
863: int i;
864:
1.43 noro 865: for ( i = 0; i < nd_wpd; i++ ) d[i] = d1[i]-d2[i];
1.1 noro 866: }
867:
1.61 noro 868: int ndl_disjoint(UINT *d1,UINT *d2)
1.1 noro 869: {
1.61 noro 870: UINT t1,t2,u,u1,u2;
1.1 noro 871: int i,j;
872:
1.65 noro 873: #if USE_UNROLL
1.1 noro 874: switch ( nd_bpe ) {
1.62 noro 875: case 3:
876: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
877: u1 = d1[i]; u2 = d2[i];
878: t1 = u1&0x38000000; t2 = u2&0x38000000; if ( t1&&t2 ) return 0;
879: t1 = u1& 0x7000000; t2 = u2& 0x7000000; if ( t1&&t2 ) return 0;
880: t1 = u1& 0xe00000; t2 = u2& 0xe00000; if ( t1&&t2 ) return 0;
881: t1 = u1& 0x1c0000; t2 = u2& 0x1c0000; if ( t1&&t2 ) return 0;
882: t1 = u1& 0x38000; t2 = u2& 0x38000; if ( t1&&t2 ) return 0;
883: t1 = u1& 0x7000; t2 = u2& 0x7000; if ( t1&&t2 ) return 0;
884: t1 = u1& 0xe00; t2 = u2& 0xe00; if ( t1&&t2 ) return 0;
885: t1 = u1& 0x1c0; t2 = u2& 0x1c0; if ( t1&&t2 ) return 0;
886: t1 = u1& 0x38; t2 = u2& 0x38; if ( t1&&t2 ) return 0;
887: t1 = u1& 0x7; t2 = u2& 0x7; if ( t1&&t2 ) return 0;
888: }
889: return 1;
890: break;
1.1 noro 891: case 4:
1.41 noro 892: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 893: u1 = d1[i]; u2 = d2[i];
894: t1 = u1&0xf0000000; t2 = u2&0xf0000000; if ( t1&&t2 ) return 0;
1.62 noro 895: t1 = u1& 0xf000000; t2 = u2& 0xf000000; if ( t1&&t2 ) return 0;
896: t1 = u1& 0xf00000; t2 = u2& 0xf00000; if ( t1&&t2 ) return 0;
897: t1 = u1& 0xf0000; t2 = u2& 0xf0000; if ( t1&&t2 ) return 0;
898: t1 = u1& 0xf000; t2 = u2& 0xf000; if ( t1&&t2 ) return 0;
899: t1 = u1& 0xf00; t2 = u2& 0xf00; if ( t1&&t2 ) return 0;
900: t1 = u1& 0xf0; t2 = u2& 0xf0; if ( t1&&t2 ) return 0;
901: t1 = u1& 0xf; t2 = u2& 0xf; if ( t1&&t2 ) return 0;
1.1 noro 902: }
903: return 1;
904: break;
905: case 6:
1.41 noro 906: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 907: u1 = d1[i]; u2 = d2[i];
908: t1 = u1&0x3f000000; t2 = u2&0x3f000000; if ( t1&&t2 ) return 0;
1.62 noro 909: t1 = u1& 0xfc0000; t2 = u2& 0xfc0000; if ( t1&&t2 ) return 0;
910: t1 = u1& 0x3f000; t2 = u2& 0x3f000; if ( t1&&t2 ) return 0;
911: t1 = u1& 0xfc0; t2 = u2& 0xfc0; if ( t1&&t2 ) return 0;
912: t1 = u1& 0x3f; t2 = u2& 0x3f; if ( t1&&t2 ) return 0;
1.1 noro 913: }
914: return 1;
915: break;
916: case 8:
1.41 noro 917: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 918: u1 = d1[i]; u2 = d2[i];
919: t1 = u1&0xff000000; t2 = u2&0xff000000; if ( t1&&t2 ) return 0;
1.62 noro 920: t1 = u1& 0xff0000; t2 = u2& 0xff0000; if ( t1&&t2 ) return 0;
921: t1 = u1& 0xff00; t2 = u2& 0xff00; if ( t1&&t2 ) return 0;
922: t1 = u1& 0xff; t2 = u2& 0xff; if ( t1&&t2 ) return 0;
1.1 noro 923: }
924: return 1;
925: break;
926: case 16:
1.41 noro 927: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 928: u1 = d1[i]; u2 = d2[i];
929: t1 = u1&0xffff0000; t2 = u2&0xffff0000; if ( t1&&t2 ) return 0;
1.62 noro 930: t1 = u1& 0xffff; t2 = u2& 0xffff; if ( t1&&t2 ) return 0;
1.1 noro 931: }
932: return 1;
933: break;
934: case 32:
1.41 noro 935: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.1 noro 936: if ( d1[i] && d2[i] ) return 0;
937: return 1;
938: break;
939: default:
1.41 noro 940: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.1 noro 941: u1 = d1[i]; u2 = d2[i];
942: for ( j = 0; j < nd_epw; j++ ) {
943: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
944: u1 >>= nd_bpe; u2 >>= nd_bpe;
945: }
946: }
947: return 1;
948: break;
949: }
1.65 noro 950: #else
951: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
952: u1 = d1[i]; u2 = d2[i];
953: for ( j = 0; j < nd_epw; j++ ) {
954: if ( (u1&nd_mask0) && (u2&nd_mask0) ) return 0;
955: u1 >>= nd_bpe; u2 >>= nd_bpe;
956: }
957: }
958: return 1;
959: #endif
1.1 noro 960: }
961:
1.61 noro 962: int ndl_check_bound2(int index,UINT *d2)
1.1 noro 963: {
1.61 noro 964: UINT u2;
965: UINT *d1;
1.5 noro 966: int i,j,ind,k;
1.1 noro 967:
1.5 noro 968: d1 = nd_bound[index];
969: ind = 0;
1.65 noro 970: #if USE_UNROLL
1.5 noro 971: switch ( nd_bpe ) {
1.62 noro 972: case 3:
973: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
974: u2 = d2[i];
975: if ( d1[ind++]+((u2>>27)&0x7) >= 0x8 ) return 1;
976: if ( d1[ind++]+((u2>>24)&0x7) >= 0x8 ) return 1;
977: if ( d1[ind++]+((u2>>21)&0x7) >= 0x8 ) return 1;
978: if ( d1[ind++]+((u2>>18)&0x7) >= 0x8 ) return 1;
979: if ( d1[ind++]+((u2>>15)&0x7) >= 0x8 ) return 1;
980: if ( d1[ind++]+((u2>>12)&0x7) >= 0x8 ) return 1;
981: if ( d1[ind++]+((u2>>9)&0x7) >= 0x8 ) return 1;
982: if ( d1[ind++]+((u2>>6)&0x7) >= 0x8 ) return 1;
983: if ( d1[ind++]+((u2>>3)&0x7) >= 0x8 ) return 1;
984: if ( d1[ind++]+(u2&0x7) >= 0x8 ) return 1;
985: }
986: return 0;
987: break;
1.5 noro 988: case 4:
1.41 noro 989: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 990: u2 = d2[i];
991: if ( d1[ind++]+((u2>>28)&0xf) >= 0x10 ) return 1;
992: if ( d1[ind++]+((u2>>24)&0xf) >= 0x10 ) return 1;
993: if ( d1[ind++]+((u2>>20)&0xf) >= 0x10 ) return 1;
994: if ( d1[ind++]+((u2>>16)&0xf) >= 0x10 ) return 1;
995: if ( d1[ind++]+((u2>>12)&0xf) >= 0x10 ) return 1;
996: if ( d1[ind++]+((u2>>8)&0xf) >= 0x10 ) return 1;
997: if ( d1[ind++]+((u2>>4)&0xf) >= 0x10 ) return 1;
998: if ( d1[ind++]+(u2&0xf) >= 0x10 ) return 1;
999: }
1000: return 0;
1001: break;
1002: case 6:
1.41 noro 1003: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1004: u2 = d2[i];
1005: if ( d1[ind++]+((u2>>24)&0x3f) >= 0x40 ) return 1;
1006: if ( d1[ind++]+((u2>>18)&0x3f) >= 0x40 ) return 1;
1007: if ( d1[ind++]+((u2>>12)&0x3f) >= 0x40 ) return 1;
1008: if ( d1[ind++]+((u2>>6)&0x3f) >= 0x40 ) return 1;
1009: if ( d1[ind++]+(u2&0x3f) >= 0x40 ) return 1;
1010: }
1011: return 0;
1012: break;
1013: case 8:
1.41 noro 1014: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1015: u2 = d2[i];
1016: if ( d1[ind++]+((u2>>24)&0xff) >= 0x100 ) return 1;
1017: if ( d1[ind++]+((u2>>16)&0xff) >= 0x100 ) return 1;
1018: if ( d1[ind++]+((u2>>8)&0xff) >= 0x100 ) return 1;
1019: if ( d1[ind++]+(u2&0xff) >= 0x100 ) return 1;
1020: }
1021: return 0;
1022: break;
1023: case 16:
1.41 noro 1024: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1025: u2 = d2[i];
1026: if ( d1[ind++]+((u2>>16)&0xffff) > 0x10000 ) return 1;
1027: if ( d1[ind++]+(u2&0xffff) > 0x10000 ) return 1;
1028: }
1029: return 0;
1030: break;
1031: case 32:
1.41 noro 1032: for ( i = nd_exporigin; i < nd_wpd; i++ )
1.5 noro 1033: if ( d1[i]+d2[i]<d1[i] ) return 1;
1034: return 0;
1035: break;
1036: default:
1.41 noro 1037: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1.5 noro 1038: u2 = d2[i];
1039: k = (nd_epw-1)*nd_bpe;
1040: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1041: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1042: }
1043: return 0;
1044: break;
1045: }
1.65 noro 1046: #else
1047: for ( i = nd_exporigin; i < nd_wpd; i++ ) {
1048: u2 = d2[i];
1049: k = (nd_epw-1)*nd_bpe;
1050: for ( j = 0; j < nd_epw; j++, k -= nd_bpe )
1051: if ( d1[ind++]+((u2>>k)&nd_mask0) > nd_mask0 ) return 1;
1052: }
1053: return 0;
1054: #endif
1.1 noro 1055: }
1056:
1.61 noro 1057: INLINE int ndl_hash_value(UINT *d)
1.1 noro 1058: {
1059: int i;
1060: int r;
1061:
1.34 noro 1062: r = 0;
1.41 noro 1063: for ( i = 0; i < nd_wpd; i++ )
1.1 noro 1064: r = ((r<<16)+d[i])%REDTAB_LEN;
1065: return r;
1066: }
1067:
1.63 noro 1068: INLINE int ndl_find_reducer(UINT *dg)
1.1 noro 1069: {
1.13 noro 1070: RHist r;
1.6 noro 1071: int d,k,i;
1.1 noro 1072:
1.63 noro 1073: d = ndl_hash_value(dg);
1.13 noro 1074: for ( r = nd_red[d], k = 0; r; r = NEXT(r), k++ ) {
1.34 noro 1075: if ( ndl_equal(dg,DL(r)) ) {
1.1 noro 1076: if ( k > 0 ) nd_notfirst++;
1077: nd_found++;
1.13 noro 1078: return r->index;
1.1 noro 1079: }
1080: }
1.13 noro 1081: if ( Reverse )
1082: for ( i = nd_psn-1; i >= 0; i-- ) {
1083: r = nd_psh[i];
1.34 noro 1084: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1085: nd_create++;
1.34 noro 1086: nd_append_red(dg,i);
1.13 noro 1087: return i;
1088: }
1089: }
1090: else
1091: for ( i = 0; i < nd_psn; i++ ) {
1092: r = nd_psh[i];
1.34 noro 1093: if ( ndl_reducible(dg,DL(r)) ) {
1.13 noro 1094: nd_create++;
1.34 noro 1095: nd_append_red(dg,i);
1.13 noro 1096: return i;
1097: }
1.1 noro 1098: }
1.6 noro 1099: return -1;
1.1 noro 1100: }
1101:
1.63 noro 1102: ND nd_merge(ND p1,ND p2)
1103: {
1104: int n,c;
1105: int t,can,td1,td2;
1106: ND r;
1107: NM m1,m2,mr0,mr,s;
1108:
1109: if ( !p1 ) return p2;
1110: else if ( !p2 ) return p1;
1111: else {
1112: can = 0;
1113: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1114: c = DL_COMPARE(DL(m1),DL(m2));
1115: switch ( c ) {
1116: case 0:
1117: s = m1; m1 = NEXT(m1);
1118: can++; NEXTNM2(mr0,mr,s);
1119: s = m2; m2 = NEXT(m2); FREENM(s);
1120: break;
1121: case 1:
1122: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1123: break;
1124: case -1:
1125: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1126: break;
1127: }
1128: }
1129: if ( !mr0 )
1130: if ( m1 ) mr0 = m1;
1131: else if ( m2 ) mr0 = m2;
1132: else return 0;
1133: else if ( m1 ) NEXT(mr) = m1;
1134: else if ( m2 ) NEXT(mr) = m2;
1135: else NEXT(mr) = 0;
1136: BDY(p1) = mr0;
1137: SG(p1) = MAX(SG(p1),SG(p2));
1138: LEN(p1) = LEN(p1)+LEN(p2)-can;
1139: FREEND(p2);
1140: return p1;
1141: }
1142: }
1143:
1.31 noro 1144: ND nd_add(int mod,ND p1,ND p2)
1.1 noro 1145: {
1146: int n,c;
1.34 noro 1147: int t,can,td1,td2;
1.1 noro 1148: ND r;
1149: NM m1,m2,mr0,mr,s;
1150:
1.34 noro 1151: if ( !p1 ) return p2;
1152: else if ( !p2 ) return p1;
1.71 noro 1153: else if ( mod == -1 ) return nd_add_sf(p1,p2);
1.34 noro 1154: else if ( !mod ) return nd_add_q(p1,p2);
1.1 noro 1155: else {
1.30 noro 1156: can = 0;
1.1 noro 1157: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1158: c = DL_COMPARE(DL(m1),DL(m2));
1.1 noro 1159: switch ( c ) {
1160: case 0:
1.19 noro 1161: t = ((CM(m1))+(CM(m2))) - mod;
1.34 noro 1162: if ( t < 0 ) t += mod;
1.1 noro 1163: s = m1; m1 = NEXT(m1);
1164: if ( t ) {
1.34 noro 1165: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1.1 noro 1166: } else {
1.34 noro 1167: can += 2; FREENM(s);
1.1 noro 1168: }
1169: s = m2; m2 = NEXT(m2); FREENM(s);
1170: break;
1171: case 1:
1172: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1173: break;
1174: case -1:
1175: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1176: break;
1177: }
1178: }
1179: if ( !mr0 )
1.34 noro 1180: if ( m1 ) mr0 = m1;
1181: else if ( m2 ) mr0 = m2;
1182: else return 0;
1183: else if ( m1 ) NEXT(mr) = m1;
1184: else if ( m2 ) NEXT(mr) = m2;
1185: else NEXT(mr) = 0;
1.1 noro 1186: BDY(p1) = mr0;
1.14 noro 1187: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1188: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.1 noro 1189: FREEND(p2);
1190: return p1;
1191: }
1192: }
1193:
1.31 noro 1194: ND nd_add_q(ND p1,ND p2)
1.17 noro 1195: {
1.30 noro 1196: int n,c,can;
1.17 noro 1197: ND r;
1198: NM m1,m2,mr0,mr,s;
1199: Q t;
1200:
1.34 noro 1201: if ( !p1 ) return p2;
1202: else if ( !p2 ) return p1;
1.31 noro 1203: else {
1.30 noro 1204: can = 0;
1.17 noro 1205: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1.34 noro 1206: c = DL_COMPARE(DL(m1),DL(m2));
1.17 noro 1207: switch ( c ) {
1208: case 0:
1209: addq(CQ(m1),CQ(m2),&t);
1210: s = m1; m1 = NEXT(m1);
1211: if ( t ) {
1.34 noro 1212: can++; NEXTNM2(mr0,mr,s); CQ(mr) = (t);
1.17 noro 1213: } else {
1.34 noro 1214: can += 2; FREENM(s);
1.17 noro 1215: }
1216: s = m2; m2 = NEXT(m2); FREENM(s);
1217: break;
1218: case 1:
1219: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1220: break;
1221: case -1:
1222: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1223: break;
1224: }
1225: }
1226: if ( !mr0 )
1.34 noro 1227: if ( m1 ) mr0 = m1;
1228: else if ( m2 ) mr0 = m2;
1229: else return 0;
1230: else if ( m1 ) NEXT(mr) = m1;
1231: else if ( m2 ) NEXT(mr) = m2;
1232: else NEXT(mr) = 0;
1.17 noro 1233: BDY(p1) = mr0;
1234: SG(p1) = MAX(SG(p1),SG(p2));
1.31 noro 1235: LEN(p1) = LEN(p1)+LEN(p2)-can;
1.17 noro 1236: FREEND(p2);
1237: return p1;
1238: }
1239: }
1240:
1.71 noro 1241: ND nd_add_sf(ND p1,ND p2)
1242: {
1243: int n,c,can;
1244: ND r;
1245: NM m1,m2,mr0,mr,s;
1246: int t;
1247:
1248: if ( !p1 ) return p2;
1249: else if ( !p2 ) return p1;
1250: else {
1251: can = 0;
1252: for ( n = NV(p1), m1 = BDY(p1), m2 = BDY(p2), mr0 = 0; m1 && m2; ) {
1253: c = DL_COMPARE(DL(m1),DL(m2));
1254: switch ( c ) {
1255: case 0:
1256: t = _addsf(CM(m1),CM(m2));
1257: s = m1; m1 = NEXT(m1);
1258: if ( t ) {
1259: can++; NEXTNM2(mr0,mr,s); CM(mr) = (t);
1260: } else {
1261: can += 2; FREENM(s);
1262: }
1263: s = m2; m2 = NEXT(m2); FREENM(s);
1264: break;
1265: case 1:
1266: s = m1; m1 = NEXT(m1); NEXTNM2(mr0,mr,s);
1267: break;
1268: case -1:
1269: s = m2; m2 = NEXT(m2); NEXTNM2(mr0,mr,s);
1270: break;
1271: }
1272: }
1273: if ( !mr0 )
1274: if ( m1 ) mr0 = m1;
1275: else if ( m2 ) mr0 = m2;
1276: else return 0;
1277: else if ( m1 ) NEXT(mr) = m1;
1278: else if ( m2 ) NEXT(mr) = m2;
1279: else NEXT(mr) = 0;
1280: BDY(p1) = mr0;
1281: SG(p1) = MAX(SG(p1),SG(p2));
1282: LEN(p1) = LEN(p1)+LEN(p2)-can;
1283: FREEND(p2);
1284: return p1;
1285: }
1286: }
1287:
1.1 noro 1288: /* ret=1 : success, ret=0 : overflow */
1.69 noro 1289: int nd_nf(int mod,ND g,NDV *ps,int full,NDC dn,ND *rp)
1.1 noro 1290: {
1.11 noro 1291: ND d;
1.1 noro 1292: NM m,mrd,tail;
1.7 noro 1293: NM mul;
1.10 noro 1294: int n,sugar,psugar,sugar0,stat,index;
1.30 noro 1295: int c,c1,c2,dummy;
1.17 noro 1296: RHist h;
1.11 noro 1297: NDV p,red;
1.69 noro 1298: Q cg,cred,gcd,tq,qq;
1.21 noro 1299: double hmag;
1.1 noro 1300:
1.69 noro 1301: if ( dn ) {
1302: if ( mod )
1303: dn->m = 1;
1304: else
1305: dn->z = ONE;
1306: }
1.1 noro 1307: if ( !g ) {
1308: *rp = 0;
1309: return 1;
1310: }
1.34 noro 1311: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.21 noro 1312:
1.14 noro 1313: sugar0 = sugar = SG(g);
1.1 noro 1314: n = NV(g);
1.61 noro 1315: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.1 noro 1316: for ( d = 0; g; ) {
1.63 noro 1317: index = ndl_find_reducer(HDL(g));
1.6 noro 1318: if ( index >= 0 ) {
1.17 noro 1319: h = nd_psh[index];
1320: ndl_sub(HDL(g),DL(h),DL(mul));
1.14 noro 1321: if ( ndl_check_bound2(index,DL(mul)) ) {
1.6 noro 1322: nd_free(g); nd_free(d);
1323: return 0;
1324: }
1.53 noro 1325: p = ps[index];
1.71 noro 1326: if ( mod == -1 )
1327: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1328: else if ( mod ) {
1.19 noro 1329: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1330: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1.16 noro 1331: } else {
1.17 noro 1332: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1.16 noro 1333: chsgnq(cg,&CQ(mul));
1.20 noro 1334: nd_mul_c_q(d,cred); nd_mul_c_q(g,cred);
1.69 noro 1335: if ( dn ) {
1336: mulq(dn->z,cred,&tq); dn->z = tq;
1337: }
1.16 noro 1338: }
1.55 noro 1339: g = nd_add(mod,g,ndv_mul_nm(mod,mul,p));
1.34 noro 1340: sugar = MAX(sugar,SG(p)+TD(DL(mul)));
1.22 noro 1341: if ( !mod && hmag && g && ((double)(p_mag((P)HCQ(g))) > hmag) ) {
1.69 noro 1342: tq = HCQ(g);
1.21 noro 1343: nd_removecont2(d,g);
1.69 noro 1344: if ( dn ) {
1345: divq(tq,HCQ(g),&qq); divq(dn->z,qq,&tq); dn->z = tq;
1346: }
1.21 noro 1347: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1348: }
1.1 noro 1349: } else if ( !full ) {
1350: *rp = g;
1351: return 1;
1352: } else {
1353: m = BDY(g);
1354: if ( NEXT(m) ) {
1.34 noro 1355: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.1 noro 1356: } else {
1357: FREEND(g); g = 0;
1358: }
1359: if ( d ) {
1.34 noro 1360: NEXT(tail)=m; tail=m; LEN(d)++;
1.1 noro 1361: } else {
1.34 noro 1362: MKND(n,m,1,d); tail = BDY(d);
1.1 noro 1363: }
1364: }
1365: }
1.34 noro 1366: if ( d ) SG(d) = sugar;
1.1 noro 1367: *rp = d;
1368: return 1;
1369: }
1.28 noro 1370:
1.53 noro 1371: int nd_nf_pbucket(int mod,ND g,NDV *ps,int full,ND *rp)
1.25 noro 1372: {
1373: int hindex,index;
1374: NDV p;
1375: ND u,d,red;
1376: NODE l;
1.31 noro 1377: NM mul,m,mrd,tail;
1.25 noro 1378: int sugar,psugar,n,h_reducible;
1379: PGeoBucket bucket;
1380: int c,c1,c2;
1.26 noro 1381: Q cg,cred,gcd,zzz;
1.25 noro 1382: RHist h;
1.28 noro 1383: double hmag,gmag;
1.25 noro 1384:
1385: if ( !g ) {
1386: *rp = 0;
1387: return 1;
1388: }
1389: sugar = SG(g);
1390: n = NV(g);
1.34 noro 1391: if ( !mod ) hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.25 noro 1392: bucket = create_pbucket();
1.31 noro 1393: add_pbucket(mod,bucket,g);
1.25 noro 1394: d = 0;
1.61 noro 1395: mul = (NM)ALLOCA(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
1.25 noro 1396: while ( 1 ) {
1.26 noro 1397: hindex = mod?head_pbucket(mod,bucket):head_pbucket_q(bucket);
1.25 noro 1398: if ( hindex < 0 ) {
1.34 noro 1399: if ( d ) SG(d) = sugar;
1.25 noro 1400: *rp = d;
1401: return 1;
1402: }
1403: g = bucket->body[hindex];
1.63 noro 1404: index = ndl_find_reducer(HDL(g));
1.25 noro 1405: if ( index >= 0 ) {
1406: h = nd_psh[index];
1407: ndl_sub(HDL(g),DL(h),DL(mul));
1408: if ( ndl_check_bound2(index,DL(mul)) ) {
1.26 noro 1409: nd_free(d);
1.25 noro 1410: free_pbucket(bucket);
1411: *rp = 0;
1412: return 0;
1413: }
1.53 noro 1414: p = ps[index];
1.71 noro 1415: if ( mod == -1 )
1416: CM(mul) = _mulsf(_invsf(HCM(p)),_chsgnsf(HCM(g)));
1417: else if ( mod ) {
1.25 noro 1418: c1 = invm(HCM(p),mod); c2 = mod-HCM(g);
1419: DMAR(c1,c2,0,mod,c); CM(mul) = c;
1420: } else {
1421: igcd_cofactor(HCQ(g),HCQ(p),&gcd,&cg,&cred);
1422: chsgnq(cg,&CQ(mul));
1.26 noro 1423: nd_mul_c_q(d,cred);
1424: mulq_pbucket(bucket,cred);
1425: g = bucket->body[hindex];
1.28 noro 1426: gmag = (double)p_mag((P)HCQ(g));
1.25 noro 1427: }
1.55 noro 1428: red = ndv_mul_nm(mod,mul,p);
1.25 noro 1429: bucket->body[hindex] = nd_remove_head(g);
1430: red = nd_remove_head(red);
1.31 noro 1431: add_pbucket(mod,bucket,red);
1.34 noro 1432: psugar = SG(p)+TD(DL(mul));
1433: sugar = MAX(sugar,psugar);
1.28 noro 1434: if ( !mod && hmag && (gmag > hmag) ) {
1435: g = normalize_pbucket(mod,bucket);
1436: if ( !g ) {
1.34 noro 1437: if ( d ) SG(d) = sugar;
1.28 noro 1438: *rp = d;
1439: return 1;
1440: }
1441: nd_removecont2(d,g);
1442: hmag = ((double)p_mag((P)HCQ(g)))*nd_scale;
1.31 noro 1443: add_pbucket(mod,bucket,g);
1.28 noro 1444: }
1.25 noro 1445: } else if ( !full ) {
1446: g = normalize_pbucket(mod,bucket);
1.34 noro 1447: if ( g ) SG(g) = sugar;
1.25 noro 1448: *rp = g;
1449: return 1;
1450: } else {
1451: m = BDY(g);
1452: if ( NEXT(m) ) {
1.34 noro 1453: BDY(g) = NEXT(m); NEXT(m) = 0; LEN(g)--;
1.25 noro 1454: } else {
1455: FREEND(g); g = 0;
1456: }
1457: bucket->body[hindex] = g;
1458: NEXT(m) = 0;
1459: if ( d ) {
1.34 noro 1460: NEXT(tail)=m; tail=m; LEN(d)++;
1.25 noro 1461: } else {
1.34 noro 1462: MKND(n,m,1,d); tail = BDY(d);
1.25 noro 1463: }
1464: }
1465: }
1466: }
1.27 noro 1467:
1.61 noro 1468: /* input : list of NDV, cand : list of NDV */
1.28 noro 1469:
1.61 noro 1470: int ndv_check_candidate(NODE input,int obpe,int oadv,EPOS oepos,NODE cand)
1.28 noro 1471: {
1472: int n,i,stat;
1473: ND nf,d;
1.61 noro 1474: NDV r;
1.45 noro 1475: NODE t,s;
1.69 noro 1476: union oNDC dn;
1.45 noro 1477:
1.61 noro 1478: ndv_setup(0,0,cand);
1.31 noro 1479: n = length(cand);
1.28 noro 1480:
1481: /* membercheck : list is a subset of Id(cand) ? */
1482: for ( t = input; t; t = NEXT(t) ) {
1.45 noro 1483: again:
1.61 noro 1484: if ( nd_bpe > obpe )
1485: r = ndv_dup_realloc((NDV)BDY(t),obpe,oadv,oepos);
1486: else
1487: r = (NDV)BDY(t);
1488: d = ndvtond(0,r);
1.69 noro 1489: stat = nd_nf(0,d,nd_ps,0,0,&nf);
1.45 noro 1490: if ( !stat ) {
1491: nd_reconstruct(0,0,0);
1492: goto again;
1493: } else if ( nf ) return 0;
1.71 noro 1494: if ( DP_Print ) { printf("."); fflush(stdout); }
1.28 noro 1495: }
1.71 noro 1496: if ( DP_Print ) { printf("\n"); }
1.28 noro 1497: /* gbcheck : cand is a GB of Id(cand) ? */
1.34 noro 1498: if ( !nd_gb(0,1) ) return 0;
1.28 noro 1499: /* XXX */
1.23 noro 1500: return 1;
1501: }
1.1 noro 1502:
1503: ND nd_remove_head(ND p)
1504: {
1505: NM m;
1506:
1507: m = BDY(p);
1508: if ( !NEXT(m) ) {
1.34 noro 1509: FREEND(p); p = 0;
1.31 noro 1510: } else {
1.34 noro 1511: BDY(p) = NEXT(m); LEN(p)--;
1.31 noro 1512: }
1.1 noro 1513: FREENM(m);
1514: return p;
1515: }
1516:
1.69 noro 1517: ND nd_separate_head(ND p,ND *head)
1518: {
1519: NM m,m0;
1520: ND r;
1521:
1522: m = BDY(p);
1523: if ( !NEXT(m) ) {
1524: *head = p; p = 0;
1525: } else {
1526: m0 = m;
1527: BDY(p) = NEXT(m); LEN(p)--;
1528: NEXT(m0) = 0;
1529: MKND(NV(p),m0,1,r);
1530: *head = r;
1531: }
1532: return p;
1533: }
1534:
1.1 noro 1535: PGeoBucket create_pbucket()
1536: {
1537: PGeoBucket g;
1538:
1539: g = CALLOC(1,sizeof(struct oPGeoBucket));
1540: g->m = -1;
1541: return g;
1542: }
1543:
1.25 noro 1544: void free_pbucket(PGeoBucket b) {
1545: int i;
1546:
1.26 noro 1547: for ( i = 0; i <= b->m; i++ )
1.25 noro 1548: if ( b->body[i] ) {
1549: nd_free(b->body[i]);
1550: b->body[i] = 0;
1551: }
1552: GC_free(b);
1553: }
1554:
1.63 noro 1555: void add_pbucket_symbolic(PGeoBucket g,ND d)
1556: {
1557: int l,i,k,m;
1558:
1559: if ( !d )
1560: return;
1561: l = LEN(d);
1562: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1563: /* 2^(k-1) < l <= 2^k (=m) */
1564: d = nd_merge(g->body[k],d);
1565: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1566: g->body[k] = 0;
1567: d = nd_merge(g->body[k+1],d);
1568: }
1569: g->body[k] = d;
1570: g->m = MAX(g->m,k);
1571: }
1572:
1.31 noro 1573: void add_pbucket(int mod,PGeoBucket g,ND d)
1.1 noro 1574: {
1.31 noro 1575: int l,i,k,m;
1.1 noro 1576:
1.31 noro 1577: if ( !d )
1578: return;
1579: l = LEN(d);
1.29 noro 1580: for ( k = 0, m = 1; l > m; k++, m <<= 1 );
1581: /* 2^(k-1) < l <= 2^k (=m) */
1.31 noro 1582: d = nd_add(mod,g->body[k],d);
1583: for ( ; d && LEN(d) > m; k++, m <<= 1 ) {
1.1 noro 1584: g->body[k] = 0;
1.31 noro 1585: d = nd_add(mod,g->body[k+1],d);
1.1 noro 1586: }
1587: g->body[k] = d;
1588: g->m = MAX(g->m,k);
1589: }
1590:
1.26 noro 1591: void mulq_pbucket(PGeoBucket g,Q c)
1592: {
1593: int k;
1594:
1595: for ( k = 0; k <= g->m; k++ )
1596: nd_mul_c_q(g->body[k],c);
1597: }
1598:
1.63 noro 1599: NM remove_head_pbucket_symbolic(PGeoBucket g)
1600: {
1601: int j,i,k,c;
1602: NM head;
1603:
1604: k = g->m;
1605: j = -1;
1606: for ( i = 0; i <= k; i++ ) {
1607: if ( !g->body[i] ) continue;
1608: if ( j < 0 ) j = i;
1609: else {
1610: c = DL_COMPARE(HDL(g->body[i]),HDL(g->body[j]));
1611: if ( c > 0 )
1612: j = i;
1613: else if ( c == 0 )
1614: g->body[i] = nd_remove_head(g->body[i]);
1615: }
1616: }
1617: if ( j < 0 ) return 0;
1618: else {
1619: head = BDY(g->body[j]);
1620: if ( !NEXT(head) ) {
1621: FREEND(g->body[j]);
1622: g->body[j] = 0;
1623: } else {
1624: BDY(g->body[j]) = NEXT(head);
1625: LEN(g->body[j])--;
1626: }
1627: return head;
1628: }
1629: }
1630:
1.19 noro 1631: int head_pbucket(int mod,PGeoBucket g)
1.1 noro 1632: {
1633: int j,i,c,k,nv,sum;
1.61 noro 1634: UINT *di,*dj;
1.1 noro 1635: ND gi,gj;
1636:
1637: k = g->m;
1638: while ( 1 ) {
1639: j = -1;
1640: for ( i = 0; i <= k; i++ ) {
1641: if ( !(gi = g->body[i]) )
1642: continue;
1643: if ( j < 0 ) {
1644: j = i;
1645: gj = g->body[j];
1646: dj = HDL(gj);
1.14 noro 1647: sum = HCM(gj);
1.1 noro 1648: } else {
1.34 noro 1649: c = DL_COMPARE(HDL(gi),dj);
1.1 noro 1650: if ( c > 0 ) {
1.34 noro 1651: if ( sum ) HCM(gj) = sum;
1652: else g->body[j] = nd_remove_head(gj);
1.1 noro 1653: j = i;
1654: gj = g->body[j];
1655: dj = HDL(gj);
1.14 noro 1656: sum = HCM(gj);
1.1 noro 1657: } else if ( c == 0 ) {
1.71 noro 1658: if ( mod == -1 )
1659: sum = _addsf(sum,HCM(gi));
1660: else {
1661: sum = sum+HCM(gi)-mod;
1662: if ( sum < 0 ) sum += mod;
1663: }
1.1 noro 1664: g->body[i] = nd_remove_head(gi);
1665: }
1666: }
1667: }
1.34 noro 1668: if ( j < 0 ) return -1;
1.1 noro 1669: else if ( sum ) {
1.14 noro 1670: HCM(gj) = sum;
1.26 noro 1671: return j;
1.31 noro 1672: } else
1.26 noro 1673: g->body[j] = nd_remove_head(gj);
1674: }
1675: }
1676:
1677: int head_pbucket_q(PGeoBucket g)
1678: {
1679: int j,i,c,k,nv;
1680: Q sum,t;
1681: ND gi,gj;
1682:
1683: k = g->m;
1684: while ( 1 ) {
1685: j = -1;
1686: for ( i = 0; i <= k; i++ ) {
1.34 noro 1687: if ( !(gi = g->body[i]) ) continue;
1.26 noro 1688: if ( j < 0 ) {
1689: j = i;
1690: gj = g->body[j];
1691: sum = HCQ(gj);
1692: } else {
1693: nv = NV(gi);
1.34 noro 1694: c = DL_COMPARE(HDL(gi),HDL(gj));
1.26 noro 1695: if ( c > 0 ) {
1.34 noro 1696: if ( sum ) HCQ(gj) = sum;
1697: else g->body[j] = nd_remove_head(gj);
1.26 noro 1698: j = i;
1699: gj = g->body[j];
1700: sum = HCQ(gj);
1701: } else if ( c == 0 ) {
1702: addq(sum,HCQ(gi),&t);
1703: sum = t;
1704: g->body[i] = nd_remove_head(gi);
1705: }
1706: }
1707: }
1.34 noro 1708: if ( j < 0 ) return -1;
1.26 noro 1709: else if ( sum ) {
1710: HCQ(gj) = sum;
1.1 noro 1711: return j;
1.31 noro 1712: } else
1.1 noro 1713: g->body[j] = nd_remove_head(gj);
1714: }
1715: }
1716:
1.25 noro 1717: ND normalize_pbucket(int mod,PGeoBucket g)
1.1 noro 1718: {
1.31 noro 1719: int i;
1.1 noro 1720: ND r,t;
1721:
1722: r = 0;
1.28 noro 1723: for ( i = 0; i <= g->m; i++ ) {
1.31 noro 1724: r = nd_add(mod,r,g->body[i]);
1.28 noro 1725: g->body[i] = 0;
1726: }
1727: g->m = -1;
1.1 noro 1728: return r;
1729: }
1730:
1.27 noro 1731: /* return value = 0 => input is not a GB */
1732:
1733: NODE nd_gb(int m,int checkonly)
1.1 noro 1734: {
1735: int i,nh,sugar,stat;
1.23 noro 1736: NODE r,g,t;
1.1 noro 1737: ND_pairs d;
1738: ND_pairs l;
1739: ND h,nf;
1.63 noro 1740: NDV nfv;
1.69 noro 1741: union oNDC dn;
1.1 noro 1742:
1.23 noro 1743: g = 0; d = 0;
1744: for ( i = 0; i < nd_psn; i++ ) {
1.1 noro 1745: d = update_pairs(d,g,i);
1746: g = update_base(g,i);
1747: }
1748: sugar = 0;
1749: while ( d ) {
1750: again:
1751: l = nd_minp(d,&d);
1.14 noro 1752: if ( SG(l) != sugar ) {
1753: sugar = SG(l);
1.71 noro 1754: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.1 noro 1755: }
1.53 noro 1756: stat = nd_sp(m,0,l,&h);
1.1 noro 1757: if ( !stat ) {
1758: NEXT(l) = d; d = l;
1.20 noro 1759: d = nd_reconstruct(m,0,d);
1.1 noro 1760: goto again;
1761: }
1.41 noro 1762: #if USE_GEOBUCKET
1.69 noro 1763: stat = m?nd_nf_pbucket(m,h,nd_ps,!Top,&nf):nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1764: #else
1.69 noro 1765: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1766: #endif
1.1 noro 1767: if ( !stat ) {
1768: NEXT(l) = d; d = l;
1.20 noro 1769: d = nd_reconstruct(m,0,d);
1.1 noro 1770: goto again;
1771: } else if ( nf ) {
1.27 noro 1772: if ( checkonly ) return 0;
1.71 noro 1773: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1774: nd_removecont(m,nf);
1775: nfv = ndtondv(m,nf); nd_free(nf);
1776: nh = ndv_newps(nfv,0);
1.1 noro 1777: d = update_pairs(d,g,nh);
1778: g = update_base(g,nh);
1779: FREENDP(l);
1780: } else {
1.71 noro 1781: if ( DP_Print ) { printf("."); fflush(stdout); }
1.1 noro 1782: FREENDP(l);
1783: }
1784: }
1.53 noro 1785: for ( t = g; t; t = NEXT(t) ) BDY(t) = (pointer)nd_ps[(int)BDY(t)];
1.1 noro 1786: return g;
1787: }
1788:
1.23 noro 1789: NODE nd_gb_trace(int m)
1.20 noro 1790: {
1791: int i,nh,sugar,stat;
1.23 noro 1792: NODE r,g,t;
1.20 noro 1793: ND_pairs d;
1794: ND_pairs l;
1795: ND h,nf,nfq;
1.63 noro 1796: NDV nfv,nfqv;
1.69 noro 1797: union oNDC dn;
1.20 noro 1798:
1.23 noro 1799: g = 0; d = 0;
1800: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 1801: d = update_pairs(d,g,i);
1802: g = update_base(g,i);
1803: }
1804: sugar = 0;
1805: while ( d ) {
1806: again:
1807: l = nd_minp(d,&d);
1808: if ( SG(l) != sugar ) {
1809: sugar = SG(l);
1.71 noro 1810: if ( DP_Print ) fprintf(asir_out,"%d",sugar);
1.20 noro 1811: }
1.53 noro 1812: stat = nd_sp(m,0,l,&h);
1.20 noro 1813: if ( !stat ) {
1814: NEXT(l) = d; d = l;
1815: d = nd_reconstruct(m,1,d);
1816: goto again;
1817: }
1.41 noro 1818: #if USE_GEOBUCKET
1.53 noro 1819: stat = nd_nf_pbucket(m,h,nd_ps,!Top,&nf);
1.41 noro 1820: #else
1.69 noro 1821: stat = nd_nf(m,h,nd_ps,!Top,0,&nf);
1.41 noro 1822: #endif
1.20 noro 1823: if ( !stat ) {
1824: NEXT(l) = d; d = l;
1825: d = nd_reconstruct(m,1,d);
1826: goto again;
1827: } else if ( nf ) {
1.70 noro 1828: if ( !nd_sp(0,1,l,&h) || !nd_nf(0,h,nd_ps_trace,!Top,0,&nfq) ) {
1829: NEXT(l) = d; d = l;
1830: d = nd_reconstruct(m,1,d);
1831: goto again;
1832: }
1.20 noro 1833: if ( nfq ) {
1.70 noro 1834: /* m|HC(nfq) => failure */
1.63 noro 1835: if ( !rem(NM(HCQ(nfq)),m) ) return 0;
1836:
1.71 noro 1837: if ( DP_Print ) { printf("+"); fflush(stdout); }
1.63 noro 1838: nd_removecont(m,nf); nfv = ndtondv(m,nf); nd_free(nf);
1839: nd_removecont(0,nfq); nfqv = ndtondv(0,nfq); nd_free(nfq);
1840: nh = ndv_newps(nfv,nfqv);
1.20 noro 1841: d = update_pairs(d,g,nh);
1842: g = update_base(g,nh);
1843: } else {
1.71 noro 1844: if ( DP_Print ) { printf("*"); fflush(stdout); }
1.20 noro 1845: }
1846: } else {
1.71 noro 1847: if ( DP_Print ) { printf("."); fflush(stdout); }
1.20 noro 1848: }
1849: FREENDP(l);
1850: }
1.23 noro 1851: for ( t = g; t; t = NEXT(t) )
1.53 noro 1852: BDY(t) = (pointer)nd_ps_trace[(int)BDY(t)];
1.20 noro 1853: return g;
1854: }
1855:
1.23 noro 1856: int ndv_compare(NDV *p1,NDV *p2)
1857: {
1.34 noro 1858: return DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1859: }
1860:
1861: int ndv_compare_rev(NDV *p1,NDV *p2)
1862: {
1.34 noro 1863: return -DL_COMPARE(HDL(*p1),HDL(*p2));
1.23 noro 1864: }
1865:
1.61 noro 1866: NODE ndv_reduceall(int m,NODE f)
1.23 noro 1867: {
1.69 noro 1868: int i,n,stat;
1869: ND nf,g,head;
1.23 noro 1870: NODE t,a0,a;
1.69 noro 1871: union oNDC dn;
1872: NDV *w;
1873: Q q;
1.23 noro 1874:
1.69 noro 1875: n = length(f);
1876: #if 0
1877: w = (NDV *)ALLOCA(n*sizeof(NDV));
1878: for ( i = 0, t = f; i < n; i++, t = NEXT(t) ) w[i] = (NDV)BDY(t);
1879: qsort(w,n,sizeof(NDV),
1880: (int (*)(const void *,const void *))ndv_compare);
1881: for ( t = f, i = 0; t; i++, t = NEXT(t) ) BDY(t) = (pointer)w[i];
1882: #endif
1883: ndv_setup(m,0,f);
1884: for ( i = 0; i < n; ) {
1885: g = ndvtond(m,nd_ps[i]);
1886: g = nd_separate_head(g,&head);
1887: stat = nd_nf(m,g,nd_ps,1,&dn,&nf);
1.50 noro 1888: if ( !stat )
1.69 noro 1889: nd_reconstruct(m,0,0);
1890: else {
1.71 noro 1891: if ( DP_Print ) { printf("."); fflush(stdout); }
1.69 noro 1892: if ( !m ) { mulq(HCQ(head),dn.z,&q); HCQ(head) = q; }
1893: nf = nd_add(m,head,nf);
1894: ndv_free(nd_ps[i]);
1.24 noro 1895: nd_removecont(m,nf);
1.69 noro 1896: nd_ps[i] = ndtondv(m,nf); nd_free(nf);
1897: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
1.50 noro 1898: i++;
1.23 noro 1899: }
1900: }
1.71 noro 1901: if ( DP_Print ) { printf("\n"); }
1.23 noro 1902: for ( a0 = 0, i = 0; i < n; i++ ) {
1903: NEXTNODE(a0,a);
1.69 noro 1904: BDY(a) = (pointer)nd_ps[i];
1.23 noro 1905: }
1906: NEXT(a) = 0;
1907: return a0;
1908: }
1909:
1.1 noro 1910: ND_pairs update_pairs( ND_pairs d, NODE /* of index */ g, int t)
1911: {
1912: ND_pairs d1,nd,cur,head,prev,remove;
1913:
1914: if ( !g ) return d;
1915: d = crit_B(d,t);
1916: d1 = nd_newpairs(g,t);
1917: d1 = crit_M(d1);
1918: d1 = crit_F(d1);
1.55 noro 1919: if ( do_weyl )
1920: head = d1;
1921: else {
1922: prev = 0; cur = head = d1;
1923: while ( cur ) {
1924: if ( crit_2( cur->i1,cur->i2 ) ) {
1925: remove = cur;
1926: if ( !prev ) head = cur = NEXT(cur);
1927: else cur = NEXT(prev) = NEXT(cur);
1928: FREENDP(remove);
1929: } else {
1930: prev = cur; cur = NEXT(cur);
1931: }
1.1 noro 1932: }
1933: }
1934: if ( !d )
1935: return head;
1936: else {
1937: nd = d;
1.34 noro 1938: while ( NEXT(nd) ) nd = NEXT(nd);
1.1 noro 1939: NEXT(nd) = head;
1940: return d;
1941: }
1942: }
1943:
1944: ND_pairs nd_newpairs( NODE g, int t )
1945: {
1946: NODE h;
1.61 noro 1947: UINT *dl;
1.34 noro 1948: int ts,s;
1.1 noro 1949: ND_pairs r,r0;
1950:
1.20 noro 1951: dl = DL(nd_psh[t]);
1.34 noro 1952: ts = SG(nd_psh[t]) - TD(dl);
1.1 noro 1953: for ( r0 = 0, h = g; h; h = NEXT(h) ) {
1954: NEXTND_pairs(r0,r);
1955: r->i1 = (int)BDY(h);
1956: r->i2 = t;
1.20 noro 1957: ndl_lcm(DL(nd_psh[r->i1]),dl,r->lcm);
1.34 noro 1958: s = SG(nd_psh[r->i1])-TD(DL(nd_psh[r->i1]));
1959: SG(r) = MAX(s,ts) + TD(LCM(r));
1.1 noro 1960: }
1961: NEXT(r) = 0;
1962: return r0;
1963: }
1964:
1965: ND_pairs crit_B( ND_pairs d, int s )
1966: {
1967: ND_pairs cur,head,prev,remove;
1.61 noro 1968: UINT *t,*tl,*lcm;
1.1 noro 1969: int td,tdl;
1970:
1971: if ( !d ) return 0;
1.20 noro 1972: t = DL(nd_psh[s]);
1.1 noro 1973: prev = 0;
1974: head = cur = d;
1.61 noro 1975: lcm = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.1 noro 1976: while ( cur ) {
1977: tl = cur->lcm;
1978: if ( ndl_reducible(tl,t)
1.20 noro 1979: && (ndl_lcm(DL(nd_psh[cur->i1]),t,lcm),!ndl_equal(lcm,tl))
1980: && (ndl_lcm(DL(nd_psh[cur->i2]),t,lcm),!ndl_equal(lcm,tl)) ) {
1.1 noro 1981: remove = cur;
1982: if ( !prev ) {
1983: head = cur = NEXT(cur);
1984: } else {
1985: cur = NEXT(prev) = NEXT(cur);
1986: }
1987: FREENDP(remove);
1988: } else {
1.34 noro 1989: prev = cur; cur = NEXT(cur);
1.1 noro 1990: }
1991: }
1992: return head;
1993: }
1994:
1995: ND_pairs crit_M( ND_pairs d1 )
1996: {
1997: ND_pairs e,d2,d3,dd,p;
1.61 noro 1998: UINT *id,*jd;
1.1 noro 1999:
2000: for ( dd = 0, e = d1; e; e = d3 ) {
2001: if ( !(d2 = NEXT(e)) ) {
2002: NEXT(e) = dd;
2003: return e;
2004: }
1.34 noro 2005: id = LCM(e);
1.1 noro 2006: for ( d3 = 0; d2; d2 = p ) {
1.34 noro 2007: p = NEXT(d2);
2008: jd = LCM(d2);
2009: if ( ndl_equal(jd,id) )
2010: ;
2011: else if ( TD(jd) > TD(id) )
1.1 noro 2012: if ( ndl_reducible(jd,id) ) continue;
2013: else ;
1.34 noro 2014: else if ( ndl_reducible(id,jd) ) goto delit;
1.1 noro 2015: NEXT(d2) = d3;
2016: d3 = d2;
2017: }
2018: NEXT(e) = dd;
2019: dd = e;
2020: continue;
2021: /**/
2022: delit: NEXT(d2) = d3;
2023: d3 = d2;
2024: for ( ; p; p = d2 ) {
2025: d2 = NEXT(p);
2026: NEXT(p) = d3;
2027: d3 = p;
2028: }
2029: FREENDP(e);
2030: }
2031: return dd;
2032: }
2033:
2034: ND_pairs crit_F( ND_pairs d1 )
2035: {
2036: ND_pairs rest, head,remove;
2037: ND_pairs last, p, r, w;
2038: int s;
2039:
2040: for ( head = last = 0, p = d1; NEXT(p); ) {
2041: r = w = equivalent_pairs(p,&rest);
1.14 noro 2042: s = SG(r);
1.1 noro 2043: w = NEXT(w);
2044: while ( w ) {
2045: if ( crit_2(w->i1,w->i2) ) {
2046: r = w;
2047: w = NEXT(w);
2048: while ( w ) {
2049: remove = w;
2050: w = NEXT(w);
2051: FREENDP(remove);
2052: }
2053: break;
1.14 noro 2054: } else if ( SG(w) < s ) {
1.1 noro 2055: FREENDP(r);
2056: r = w;
1.14 noro 2057: s = SG(r);
1.1 noro 2058: w = NEXT(w);
2059: } else {
2060: remove = w;
2061: w = NEXT(w);
2062: FREENDP(remove);
2063: }
2064: }
2065: if ( last ) NEXT(last) = r;
2066: else head = r;
2067: NEXT(last = r) = 0;
2068: p = rest;
2069: if ( !p ) return head;
2070: }
2071: if ( !last ) return p;
2072: NEXT(last) = p;
2073: return head;
2074: }
2075:
2076: int crit_2( int dp1, int dp2 )
2077: {
1.20 noro 2078: return ndl_disjoint(DL(nd_psh[dp1]),DL(nd_psh[dp2]));
1.1 noro 2079: }
2080:
1.40 noro 2081: ND_pairs equivalent_pairs( ND_pairs d1, ND_pairs *prest )
1.1 noro 2082: {
2083: ND_pairs w,p,r,s;
1.61 noro 2084: UINT *d;
1.1 noro 2085:
2086: w = d1;
1.34 noro 2087: d = LCM(w);
1.1 noro 2088: s = NEXT(w);
2089: NEXT(w) = 0;
2090: for ( r = 0; s; s = p ) {
2091: p = NEXT(s);
1.34 noro 2092: if ( ndl_equal(d,LCM(s)) ) {
1.39 noro 2093: NEXT(s) = w; w = s;
1.1 noro 2094: } else {
1.39 noro 2095: NEXT(s) = r; r = s;
1.1 noro 2096: }
2097: }
2098: *prest = r;
2099: return w;
2100: }
2101:
2102: NODE update_base(NODE nd,int ndp)
2103: {
1.61 noro 2104: UINT *dl, *dln;
1.1 noro 2105: NODE last, p, head;
2106:
1.20 noro 2107: dl = DL(nd_psh[ndp]);
1.1 noro 2108: for ( head = last = 0, p = nd; p; ) {
1.20 noro 2109: dln = DL(nd_psh[(int)BDY(p)]);
1.34 noro 2110: if ( ndl_reducible( dln, dl ) ) {
1.1 noro 2111: p = NEXT(p);
2112: if ( last ) NEXT(last) = p;
2113: } else {
2114: if ( !last ) head = p;
2115: p = NEXT(last = p);
2116: }
2117: }
2118: head = append_one(head,ndp);
2119: return head;
2120: }
2121:
2122: ND_pairs nd_minp( ND_pairs d, ND_pairs *prest )
2123: {
2124: ND_pairs m,ml,p,l;
1.61 noro 2125: UINT *lcm;
1.33 noro 2126: int s,td,len,tlen,c,c1;
1.1 noro 2127:
2128: if ( !(p = NEXT(m = d)) ) {
2129: *prest = p;
2130: NEXT(m) = 0;
2131: return m;
2132: }
1.14 noro 2133: s = SG(m);
1.33 noro 2134: for ( ml = 0, l = m; p; p = NEXT(l = p) )
1.34 noro 2135: if ( (SG(p) < s)
2136: || ((SG(p) == s) && (DL_COMPARE(LCM(p),LCM(m)) < 0)) ) {
1.39 noro 2137: ml = l; m = p; s = SG(m);
1.1 noro 2138: }
2139: if ( !ml ) *prest = NEXT(m);
2140: else {
2141: NEXT(ml) = NEXT(m);
2142: *prest = d;
2143: }
2144: NEXT(m) = 0;
2145: return m;
2146: }
2147:
1.63 noro 2148: ND_pairs nd_minsugarp( ND_pairs d, ND_pairs *prest )
2149: {
1.69 noro 2150: int msugar,i;
1.63 noro 2151: ND_pairs t,dm0,dm,dr0,dr;
2152:
2153: for ( msugar = SG(d), t = NEXT(d); t; t = NEXT(t) )
2154: if ( SG(t) < msugar ) msugar = SG(t);
2155: dm0 = 0; dr0 = 0;
1.69 noro 2156: for ( i = 0, t = d; t; t = NEXT(t) )
2157: if ( i < nd_f4_nsp && SG(t) == msugar ) {
1.63 noro 2158: if ( dm0 ) NEXT(dm) = t;
2159: else dm0 = t;
2160: dm = t;
1.69 noro 2161: i++;
1.63 noro 2162: } else {
2163: if ( dr0 ) NEXT(dr) = t;
2164: else dr0 = t;
2165: dr = t;
2166: }
2167: NEXT(dm) = 0;
2168: if ( dr0 ) NEXT(dr) = 0;
2169: *prest = dr0;
2170: return dm0;
2171: }
2172:
2173: int ndv_newps(NDV a,NDV aq)
1.1 noro 2174: {
1.3 noro 2175: int len;
1.13 noro 2176: RHist r;
1.20 noro 2177: NDV b;
1.3 noro 2178:
1.1 noro 2179: if ( nd_psn == nd_pslen ) {
2180: nd_pslen *= 2;
1.11 noro 2181: nd_ps = (NDV *)REALLOC((char *)nd_ps,nd_pslen*sizeof(NDV));
1.53 noro 2182: nd_ps_trace = (NDV *)REALLOC((char *)nd_ps_trace,nd_pslen*sizeof(NDV));
1.13 noro 2183: nd_psh = (RHist *)REALLOC((char *)nd_psh,nd_pslen*sizeof(RHist));
1.61 noro 2184: nd_bound = (UINT **)
2185: REALLOC((char *)nd_bound,nd_pslen*sizeof(UINT *));
1.1 noro 2186: }
1.39 noro 2187: NEWRHist(r); nd_psh[nd_psn] = r;
1.63 noro 2188: nd_ps[nd_psn] = a;
1.39 noro 2189: if ( aq ) {
1.63 noro 2190: nd_ps_trace[nd_psn] = aq;
2191: nd_bound[nd_psn] = ndv_compute_bound(aq);
1.39 noro 2192: SG(r) = SG(aq); ndl_copy(HDL(aq),DL(r));
1.53 noro 2193: } else {
1.63 noro 2194: nd_bound[nd_psn] = ndv_compute_bound(a);
1.53 noro 2195: SG(r) = SG(a); ndl_copy(HDL(a),DL(r));
1.39 noro 2196: }
1.1 noro 2197: return nd_psn++;
2198: }
2199:
1.61 noro 2200: void ndv_setup(int mod,int trace,NODE f)
1.1 noro 2201: {
1.5 noro 2202: int i,j,td,len,max;
1.1 noro 2203: NODE s,s0,f0;
1.61 noro 2204: UINT *d;
1.13 noro 2205: RHist r;
1.69 noro 2206: NDV *w;
1.61 noro 2207: NDV a,am;
1.11 noro 2208:
2209: nd_found = 0; nd_notfirst = 0; nd_create = 0;
1.1 noro 2210:
1.69 noro 2211: for ( nd_psn = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) nd_psn++;
2212: w = (NDV *)ALLOCA(nd_psn*sizeof(NDV));
2213: for ( i = 0, s = f; s; s = NEXT(s) ) if ( BDY(s) ) w[i++] = BDY(s);
2214: qsort(w,nd_psn,sizeof(NDV),
2215: (int (*)(const void *,const void *))ndv_compare);
2216: nd_pslen = 2*nd_psn;
1.11 noro 2217: nd_ps = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.53 noro 2218: nd_ps_trace = (NDV *)MALLOC(nd_pslen*sizeof(NDV));
1.13 noro 2219: nd_psh = (RHist *)MALLOC(nd_pslen*sizeof(RHist));
1.61 noro 2220: nd_bound = (UINT **)MALLOC(nd_pslen*sizeof(UINT *));
1.57 noro 2221:
1.11 noro 2222: if ( !nd_red )
1.13 noro 2223: nd_red = (RHist *)MALLOC(REDTAB_LEN*sizeof(RHist));
1.71 noro 2224: for ( i = 0; i < REDTAB_LEN; i++ ) nd_red[i] = 0;
1.69 noro 2225: for ( i = 0; i < nd_psn; i++ ) {
1.39 noro 2226: if ( trace ) {
1.69 noro 2227: a = nd_ps_trace[i] = ndv_dup(0,w[i]);
1.61 noro 2228: ndv_removecont(0,a);
2229: am = nd_ps[i] = ndv_dup(mod,a);
2230: ndv_mod(mod,am);
2231: ndv_removecont(mod,am);
2232: } else {
1.69 noro 2233: a = nd_ps[i] = ndv_dup(mod,w[i]);
1.61 noro 2234: ndv_removecont(mod,a);
1.39 noro 2235: }
1.61 noro 2236: NEWRHist(r); SG(r) = HTD(a); ndl_copy(HDL(a),DL(r));
1.57 noro 2237: nd_bound[i] = ndv_compute_bound(a);
1.20 noro 2238: nd_psh[i] = r;
2239: }
2240: }
2241:
1.63 noro 2242: void nd_gr(LIST f,LIST v,int m,int f4,struct order_spec *ord,LIST *rp)
1.1 noro 2243: {
1.61 noro 2244: VL tv,fv,vv,vc;
1.1 noro 2245: NODE fd,fd0,r,r0,t,x,s,xx;
1.61 noro 2246: int e,max,nvar;
1.69 noro 2247: NDV b;
1.1 noro 2248:
1.74 ! noro 2249: ndv_alloc = 0;
1.1 noro 2250: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 ! noro 2251: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.32 noro 2252: nd_init_ord(ord);
1.61 noro 2253: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2254: for ( tv = vv; tv; tv = NEXT(tv) ) {
2255: e = getdeg(tv->v,(P)BDY(t));
2256: max = MAX(e,max);
2257: }
2258: nd_setup_parameters(nvar,max);
1.1 noro 2259: for ( fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
1.61 noro 2260: b = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 2261: if ( m ) ndv_mod(m,b);
1.61 noro 2262: if ( b ) { NEXTNODE(fd0,fd); BDY(fd) = (pointer)b; }
1.1 noro 2263: }
2264: if ( fd0 ) NEXT(fd) = 0;
1.61 noro 2265: ndv_setup(m,0,fd0);
1.63 noro 2266: x = f4?nd_f4(m):nd_gb(m,0);
1.61 noro 2267: x = ndv_reducebase(x);
2268: x = ndv_reduceall(m,x);
1.23 noro 2269: for ( r0 = 0, t = x; t; t = NEXT(t) ) {
1.1 noro 2270: NEXTNODE(r0,r);
1.61 noro 2271: BDY(r) = ndvtop(m,CO,vv,BDY(t));
1.20 noro 2272: }
2273: if ( r0 ) NEXT(r) = 0;
2274: MKLIST(*rp,r0);
1.74 ! noro 2275: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.20 noro 2276: }
2277:
1.52 noro 2278: void nd_gr_trace(LIST f,LIST v,int trace,int homo,struct order_spec *ord,LIST *rp)
1.20 noro 2279: {
2280: struct order_spec ord1;
1.61 noro 2281: VL tv,fv,vv,vc;
1.27 noro 2282: NODE fd,fd0,in0,in,r,r0,t,s,cand;
1.61 noro 2283: int m,nocheck,nvar,mindex,e,max;
2284: NDV c;
2285: NMV a;
1.27 noro 2286: P p;
1.61 noro 2287: EPOS oepos;
2288: int obpe,oadv,wmax,i,len,cbpe;
1.20 noro 2289:
2290: get_vars((Obj)f,&fv); pltovl(v,&vv);
1.74 ! noro 2291: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.52 noro 2292: nocheck = 0;
2293: mindex = 0;
2294:
2295: /* setup modulus */
2296: if ( trace < 0 ) {
2297: trace = -trace;
2298: nocheck = 1;
2299: }
2300: m = trace > 1 ? trace : get_lprime(mindex);
1.61 noro 2301: for ( t = BDY(f), max = 0; t; t = NEXT(t) )
2302: for ( tv = vv; tv; tv = NEXT(tv) ) {
2303: e = getdeg(tv->v,(P)BDY(t));
2304: max = MAX(e,max);
1.23 noro 2305: }
1.61 noro 2306: nd_init_ord(ord);
2307: nd_setup_parameters(nvar,max);
2308: obpe = nd_bpe; oadv = nmv_adv; oepos = nd_epos;
2309: for ( in0 = 0, fd0 = 0, t = BDY(f); t; t = NEXT(t) ) {
2310: c = ptondv(CO,vv,(P)BDY(t));
2311: if ( c ) {
2312: NEXTNODE(in0,in); BDY(in) = (pointer)c;
2313: NEXTNODE(fd0,fd); BDY(fd) = (pointer)ndv_dup(0,c);
1.23 noro 2314: }
1.61 noro 2315: }
2316: if ( in0 ) NEXT(in) = 0;
2317: if ( fd0 ) NEXT(fd) = 0;
2318: if ( homo ) {
2319: for ( t = in0, wmax = 0; t; t = NEXT(t) ) {
2320: c = (NDV)BDY(t); len = LEN(c);
2321: for ( a = BDY(c), i = 0; i < len; i++, NMV_ADV(a) )
2322: wmax = MAX(TD(DL(a)),wmax);
2323: }
2324: homogenize_order(ord,nvar,&ord1);
2325: nd_init_ord(&ord1);
2326: nd_setup_parameters(nvar+1,wmax);
2327: for ( t = fd0; t; t = NEXT(t) )
2328: ndv_homogenize((NDV)BDY(t),obpe,oadv,oepos);
1.27 noro 2329: }
1.52 noro 2330: while ( 1 ) {
1.61 noro 2331: ndv_setup(m,1,fd0);
1.27 noro 2332: cand = nd_gb_trace(m);
1.52 noro 2333: if ( !cand ) {
2334: /* failure */
1.61 noro 2335: if ( trace > 1 ) { *rp = 0; return; }
2336: else m = get_lprime(++mindex);
1.52 noro 2337: continue;
2338: }
1.27 noro 2339: if ( homo ) {
2340: /* dehomogenization */
1.61 noro 2341: for ( t = cand; t; t = NEXT(t) ) ndv_dehomogenize((NDV)BDY(t),ord);
1.45 noro 2342: nd_init_ord(ord);
1.61 noro 2343: nd_setup_parameters(nvar,0);
1.27 noro 2344: }
1.61 noro 2345: cand = ndv_reducebase(cand);
2346: cand = ndv_reduceall(0,cand);
2347: if ( nocheck )
2348: break;
2349: cbpe = nd_bpe;
2350: if ( ndv_check_candidate(in0,obpe,oadv,oepos,cand) )
1.52 noro 2351: /* success */
2352: break;
2353: else if ( trace > 1 ) {
2354: /* failure */
2355: *rp = 0; return;
1.61 noro 2356: } else {
1.52 noro 2357: /* try the next modulus */
2358: m = get_lprime(++mindex);
1.61 noro 2359: /* reset the parameters */
2360: if ( homo ) {
2361: nd_init_ord(&ord1);
2362: nd_setup_parameters(nvar+1,wmax);
2363: } else {
2364: nd_init_ord(ord);
2365: nd_setup_parameters(nvar,max);
2366: }
2367: }
1.52 noro 2368: }
1.27 noro 2369: /* dp->p */
1.61 noro 2370: nd_bpe = cbpe;
1.73 noro 2371: nd_setup_parameters(nd_nvar,0);
1.61 noro 2372: for ( r = cand; r; r = NEXT(r) ) BDY(r) = (pointer)ndvtop(0,CO,vv,BDY(r));
1.27 noro 2373: MKLIST(*rp,cand);
1.1 noro 2374: }
2375:
1.61 noro 2376: void dltondl(int n,DL dl,UINT *r)
1.1 noro 2377: {
1.61 noro 2378: UINT *d;
1.57 noro 2379: int i,j,l,s,ord_l;
1.43 noro 2380: struct order_pair *op;
1.1 noro 2381:
2382: d = dl->d;
1.41 noro 2383: for ( i = 0; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2384: if ( nd_blockmask ) {
2385: l = nd_blockmask->n;
2386: op = nd_blockmask->order_pair;
2387: for ( j = 0, s = 0; j < l; j++ ) {
2388: ord_l = op[j].length;
1.57 noro 2389: for ( i = 0; i < ord_l; i++, s++ ) PUT_EXP(r,s,d[s]);
1.43 noro 2390: }
2391: TD(r) = ndl_weight(r);
1.61 noro 2392: ndl_weight_mask(r);
1.43 noro 2393: } else {
1.56 noro 2394: for ( i = 0; i < n; i++ ) PUT_EXP(r,i,d[i]);
1.43 noro 2395: TD(r) = ndl_weight(r);
2396: }
1.1 noro 2397: }
2398:
1.61 noro 2399: DL ndltodl(int n,UINT *ndl)
1.1 noro 2400: {
2401: DL dl;
2402: int *d;
1.57 noro 2403: int i,j,l,s,ord_l;
1.43 noro 2404: struct order_pair *op;
1.1 noro 2405:
2406: NEWDL(dl,n);
1.34 noro 2407: dl->td = TD(ndl);
1.1 noro 2408: d = dl->d;
1.43 noro 2409: if ( nd_blockmask ) {
2410: l = nd_blockmask->n;
2411: op = nd_blockmask->order_pair;
2412: for ( j = 0, s = 0; j < l; j++ ) {
2413: ord_l = op[j].length;
1.57 noro 2414: for ( i = 0; i < ord_l; i++, s++ ) d[s] = GET_EXP(ndl,s);
1.43 noro 2415: }
2416: } else {
1.56 noro 2417: for ( i = 0; i < n; i++ ) d[i] = GET_EXP(ndl,i);
1.43 noro 2418: }
1.1 noro 2419: return dl;
2420: }
2421:
1.61 noro 2422: void ndl_print(UINT *dl)
1.1 noro 2423: {
2424: int n;
1.57 noro 2425: int i,j,l,ord_l,s,s0;
1.43 noro 2426: struct order_pair *op;
1.1 noro 2427:
2428: n = nd_nvar;
2429: printf("<<");
1.43 noro 2430: if ( nd_blockmask ) {
2431: l = nd_blockmask->n;
2432: op = nd_blockmask->order_pair;
2433: for ( j = 0, s = s0 = 0; j < l; j++ ) {
2434: ord_l = op[j].length;
1.57 noro 2435: for ( i = 0; i < ord_l; i++, s++ )
2436: printf(s==n-1?"%d":"%d,",GET_EXP(dl,s));
1.43 noro 2437: }
2438: } else {
1.56 noro 2439: for ( i = 0; i < n; i++ ) printf(i==n-1?"%d":"%d,",GET_EXP(dl,i));
1.43 noro 2440: }
1.1 noro 2441: printf(">>");
2442: }
2443:
2444: void nd_print(ND p)
2445: {
2446: NM m;
2447:
2448: if ( !p )
2449: printf("0\n");
2450: else {
2451: for ( m = BDY(p); m; m = NEXT(m) ) {
1.71 noro 2452: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
2453: else printf("+%d*",CM(m));
1.14 noro 2454: ndl_print(DL(m));
1.1 noro 2455: }
2456: printf("\n");
2457: }
2458: }
2459:
1.16 noro 2460: void nd_print_q(ND p)
2461: {
2462: NM m;
2463:
2464: if ( !p )
2465: printf("0\n");
2466: else {
2467: for ( m = BDY(p); m; m = NEXT(m) ) {
2468: printf("+");
1.74 ! noro 2469: printexpr(CO,(Obj)CQ(m));
1.16 noro 2470: printf("*");
2471: ndl_print(DL(m));
2472: }
2473: printf("\n");
2474: }
2475: }
2476:
1.1 noro 2477: void ndp_print(ND_pairs d)
2478: {
2479: ND_pairs t;
2480:
1.34 noro 2481: for ( t = d; t; t = NEXT(t) ) printf("%d,%d ",t->i1,t->i2);
1.1 noro 2482: printf("\n");
2483: }
2484:
1.20 noro 2485: void nd_removecont(int mod,ND p)
1.16 noro 2486: {
2487: int i,n;
2488: Q *w;
2489: Q dvr,t;
2490: NM m;
1.21 noro 2491: struct oVECT v;
2492: N q,r;
1.16 noro 2493:
1.71 noro 2494: if ( mod == -1 ) nd_mul_c(mod,p,_invsf(HCM(p)));
2495: else if ( mod ) nd_mul_c(mod,p,invm(HCM(p),mod));
1.20 noro 2496: else {
2497: for ( m = BDY(p), n = 0; m; m = NEXT(m), n++ );
2498: w = (Q *)ALLOCA(n*sizeof(Q));
1.21 noro 2499: v.len = n;
2500: v.body = (pointer *)w;
1.34 noro 2501: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2502: removecont_array(w,n);
2503: for ( m = BDY(p), i = 0; i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
1.16 noro 2504: }
2505: }
2506:
1.21 noro 2507: void nd_removecont2(ND p1,ND p2)
2508: {
2509: int i,n1,n2,n;
2510: Q *w;
2511: Q dvr,t;
2512: NM m;
2513: struct oVECT v;
2514: N q,r;
2515:
2516: if ( !p1 ) {
2517: nd_removecont(0,p2); return;
2518: } else if ( !p2 ) {
2519: nd_removecont(0,p1); return;
2520: }
2521: n1 = nd_length(p1);
2522: n2 = nd_length(p2);
2523: n = n1+n2;
2524: w = (Q *)ALLOCA(n*sizeof(Q));
2525: v.len = n;
2526: v.body = (pointer *)w;
1.34 noro 2527: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) w[i] = CQ(m);
2528: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) w[i] = CQ(m);
1.21 noro 2529: removecont_array(w,n);
2530: for ( m = BDY(p1), i = 0; i < n1; m = NEXT(m), i++ ) CQ(m) = w[i];
2531: for ( m = BDY(p2); i < n; m = NEXT(m), i++ ) CQ(m) = w[i];
2532: }
2533:
1.20 noro 2534: void ndv_removecont(int mod,NDV p)
1.16 noro 2535: {
2536: int i,len;
2537: Q *w;
2538: Q dvr,t;
2539: NMV m;
2540:
1.71 noro 2541: if ( mod == -1 )
2542: ndv_mul_c(mod,p,_invsf(HCM(p)));
2543: else if ( mod )
1.20 noro 2544: ndv_mul_c(mod,p,invm(HCM(p),mod));
2545: else {
2546: len = p->len;
2547: w = (Q *)ALLOCA(len*sizeof(Q));
1.34 noro 2548: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) w[i] = CQ(m);
1.20 noro 2549: sortbynm(w,len);
2550: qltozl(w,len,&dvr);
2551: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ ) {
2552: divq(CQ(m),dvr,&t); CQ(m) = t;
2553: }
1.16 noro 2554: }
1.21 noro 2555: }
2556:
1.61 noro 2557: void ndv_homogenize(NDV p,int obpe,int oadv,EPOS oepos)
2558: {
2559: int len,i,max;
2560: NMV m,mr0,mr,t;
2561:
2562: len = p->len;
2563: for ( m = BDY(p), i = 0, max = 0; i < len; NMV_OADV(m), i++ )
2564: max = MAX(max,TD(DL(m)));
2565: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
2566: m = (NMV)((char *)mr0+(len-1)*oadv);
2567: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
2568: t = (NMV)ALLOCA(nmv_adv);
2569: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
2570: ndl_homogenize(DL(m),DL(t),obpe,oepos,max);
2571: CQ(mr) = CQ(m);
2572: ndl_copy(DL(t),DL(mr));
2573: }
2574: NV(p)++;
2575: BDY(p) = mr0;
2576: }
2577:
1.45 noro 2578: void ndv_dehomogenize(NDV p,struct order_spec *ord)
1.23 noro 2579: {
1.45 noro 2580: int i,j,adj,len,newnvar,newwpd,newadv,newexporigin;
1.23 noro 2581: Q *w;
2582: Q dvr,t;
2583: NMV m,r;
2584:
2585: len = p->len;
2586: newnvar = nd_nvar-1;
1.48 noro 2587: newexporigin = nd_get_exporigin(ord);
1.45 noro 2588: newwpd = newnvar/nd_epw+(newnvar%nd_epw?1:0)+newexporigin;
1.23 noro 2589: for ( m = BDY(p), i = 0; i < len; NMV_ADV(m), i++ )
1.34 noro 2590: ndl_dehomogenize(DL(m));
1.23 noro 2591: if ( newwpd != nd_wpd ) {
1.61 noro 2592: newadv = sizeof(struct oNMV)+(newwpd-1)*sizeof(UINT);
2593: for ( m = r = BDY(p), i = 0; i < len; NMV_ADV(m), NDV_NADV(r), i++ ) {
1.45 noro 2594: CQ(r) = CQ(m);
2595: for ( j = 0; j < newexporigin; j++ ) DL(r)[j] = DL(m)[j];
2596: adj = nd_exporigin-newexporigin;
2597: for ( ; j < newwpd; j++ ) DL(r)[j] = DL(m)[j+adj];
1.23 noro 2598: }
2599: }
2600: NV(p)--;
2601: }
2602:
1.21 noro 2603: void removecont_array(Q *c,int n)
2604: {
2605: struct oVECT v;
2606: Q d0,d1,a,u,u1,gcd;
2607: int i;
2608: N qn,rn,gn;
2609: Q *q,*r;
2610:
2611: q = (Q *)ALLOCA(n*sizeof(Q));
2612: r = (Q *)ALLOCA(n*sizeof(Q));
2613: v.id = O_VECT; v.len = n; v.body = (pointer *)c;
2614: igcdv_estimate(&v,&d0);
2615: for ( i = 0; i < n; i++ ) {
2616: divn(NM(c[i]),NM(d0),&qn,&rn);
2617: NTOQ(qn,SGN(c[i])*SGN(d0),q[i]);
2618: NTOQ(rn,SGN(c[i]),r[i]);
2619: }
1.34 noro 2620: for ( i = 0; i < n; i++ ) if ( r[i] ) break;
1.21 noro 2621: if ( i < n ) {
2622: v.id = O_VECT; v.len = n; v.body = (pointer *)r;
2623: igcdv(&v,&d1);
2624: gcdn(NM(d0),NM(d1),&gn); NTOQ(gn,1,gcd);
2625: divsn(NM(d0),gn,&qn); NTOQ(qn,1,a);
2626: for ( i = 0; i < n; i++ ) {
2627: mulq(a,q[i],&u);
2628: if ( r[i] ) {
2629: divsn(NM(r[i]),gn,&qn); NTOQ(qn,SGN(r[i]),u1);
2630: addq(u,u1,&q[i]);
2631: } else
2632: q[i] = u;
2633: }
2634: }
1.34 noro 2635: for ( i = 0; i < n; i++ ) c[i] = q[i];
1.16 noro 2636: }
2637:
1.19 noro 2638: void nd_mul_c(int mod,ND p,int mul)
1.1 noro 2639: {
2640: NM m;
2641: int c,c1;
2642:
1.34 noro 2643: if ( !p ) return;
1.71 noro 2644: if ( mod == -1 )
2645: for ( m = BDY(p); m; m = NEXT(m) )
2646: CM(m) = _mulsf(CM(m),mul);
2647: else
2648: for ( m = BDY(p); m; m = NEXT(m) ) {
2649: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2650: }
1.1 noro 2651: }
2652:
1.16 noro 2653: void nd_mul_c_q(ND p,Q mul)
2654: {
2655: NM m;
2656: Q c;
2657:
1.34 noro 2658: if ( !p ) return;
1.16 noro 2659: for ( m = BDY(p); m; m = NEXT(m) ) {
2660: mulq(CQ(m),mul,&c); CQ(m) = c;
2661: }
2662: }
2663:
1.61 noro 2664: void nd_mul_c_p(VL vl,ND p,P mul)
2665: {
2666: NM m;
2667: P c;
2668:
2669: if ( !p ) return;
2670: for ( m = BDY(p); m; m = NEXT(m) ) {
2671: mulp(vl,CP(m),mul,&c); CP(m) = c;
2672: }
2673: }
2674:
1.1 noro 2675: void nd_free(ND p)
2676: {
2677: NM t,s;
2678:
1.34 noro 2679: if ( !p ) return;
1.1 noro 2680: t = BDY(p);
2681: while ( t ) {
2682: s = NEXT(t);
2683: FREENM(t);
2684: t = s;
2685: }
2686: FREEND(p);
2687: }
2688:
1.23 noro 2689: void ndv_free(NDV p)
2690: {
2691: GC_free(BDY(p));
2692: }
2693:
1.61 noro 2694: void nd_append_red(UINT *d,int i)
1.1 noro 2695: {
1.13 noro 2696: RHist m,m0;
1.1 noro 2697: int h;
2698:
1.13 noro 2699: NEWRHist(m);
1.34 noro 2700: h = ndl_hash_value(d);
1.13 noro 2701: m->index = i;
1.14 noro 2702: ndl_copy(d,DL(m));
1.1 noro 2703: NEXT(m) = nd_red[h];
2704: nd_red[h] = m;
2705: }
2706:
1.61 noro 2707: UINT *ndv_compute_bound(NDV p)
1.1 noro 2708: {
1.61 noro 2709: UINT *d1,*d2,*t;
2710: UINT u;
1.57 noro 2711: int i,j,k,l,len,ind;
1.45 noro 2712: NMV m;
1.1 noro 2713:
2714: if ( !p )
2715: return 0;
1.61 noro 2716: d1 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
2717: d2 = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.45 noro 2718: len = LEN(p);
2719: m = BDY(p); ndl_copy(DL(m),d1); NMV_ADV(m);
2720: for ( i = 1; i < len; i++, NMV_ADV(m) ) {
1.14 noro 2721: ndl_lcm(DL(m),d1,d2);
1.1 noro 2722: t = d1; d1 = d2; d2 = t;
2723: }
1.12 noro 2724: l = nd_nvar+31;
1.61 noro 2725: t = (UINT *)MALLOC_ATOMIC(l*sizeof(UINT));
1.57 noro 2726: for ( i = nd_exporigin, ind = 0; i < nd_wpd; i++ ) {
2727: u = d1[i];
2728: k = (nd_epw-1)*nd_bpe;
2729: for ( j = 0; j < nd_epw; j++, k -= nd_bpe, ind++ )
2730: t[ind] = (u>>k)&nd_mask0;
2731: }
2732: for ( ; ind < l; ind++ ) t[ind] = 0;
1.1 noro 2733: return t;
2734: }
2735:
1.48 noro 2736: int nd_get_exporigin(struct order_spec *ord)
2737: {
1.51 noro 2738: switch ( ord->id ) {
1.41 noro 2739: case 0:
1.48 noro 2740: return 1;
1.41 noro 2741: case 1:
2742: /* block order */
1.43 noro 2743: /* d[0]:weight d[1]:w0,...,d[nd_exporigin-1]:w(n-1) */
1.48 noro 2744: return ord->ord.block.length+1;
1.41 noro 2745: case 2:
1.52 noro 2746: error("nd_get_exporigin : matrix order is not supported yet.");
1.41 noro 2747: }
1.48 noro 2748: }
2749:
1.61 noro 2750: void nd_setup_parameters(int nvar,int max) {
1.57 noro 2751: int i,j,n,elen,ord_o,ord_l,l,s;
2752: struct order_pair *op;
1.73 noro 2753: int bpe;
1.48 noro 2754:
1.73 noro 2755: if ( !max ) bpe = nd_bpe;
2756: else if ( max < 2 ) bpe = 1;
2757: else if ( max < 4 ) bpe = 2;
2758: else if ( max < 8 ) bpe = 3;
2759: else if ( max < 16 ) bpe = 4;
2760: else if ( max < 32 ) bpe = 5;
2761: else if ( max < 64 ) bpe = 6;
2762: else if ( max < 256 ) bpe = 8;
2763: else if ( max < 1024 ) bpe = 10;
2764: else if ( max < 65536 ) bpe = 16;
2765: else bpe = 32;
2766: if ( bpe != nd_bpe || nvar != nd_nvar )
2767: nd_free_private_storage();
2768: nd_bpe = bpe;
2769: nd_nvar = nvar;
1.61 noro 2770: nd_epw = (sizeof(UINT)*8)/nd_bpe;
1.48 noro 2771: elen = nd_nvar/nd_epw+(nd_nvar%nd_epw?1:0);
2772:
2773: nd_exporigin = nd_get_exporigin(nd_ord);
1.43 noro 2774: nd_wpd = nd_exporigin+elen;
1.57 noro 2775:
1.1 noro 2776: if ( nd_bpe < 32 ) {
2777: nd_mask0 = (1<<nd_bpe)-1;
2778: } else {
2779: nd_mask0 = 0xffffffff;
2780: }
2781: bzero(nd_mask,sizeof(nd_mask));
2782: nd_mask1 = 0;
2783: for ( i = 0; i < nd_epw; i++ ) {
2784: nd_mask[nd_epw-i-1] = (nd_mask0<<(i*nd_bpe));
2785: nd_mask1 |= (1<<(nd_bpe-1))<<(i*nd_bpe);
2786: }
1.61 noro 2787: nm_adv = sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT);
2788: nmv_adv = sizeof(struct oNMV)+(nd_wpd-1)*sizeof(UINT);
1.57 noro 2789: nd_epos = nd_create_epos(nd_ord);
1.43 noro 2790: nd_blockmask = nd_create_blockmask(nd_ord);
1.1 noro 2791: }
2792:
1.20 noro 2793: ND_pairs nd_reconstruct(int mod,int trace,ND_pairs d)
1.1 noro 2794: {
1.37 noro 2795: int i,obpe,oadv,h;
1.71 noro 2796: static NM prev_nm_free_list;
2797: static ND_pairs prev_ndp_free_list;
1.13 noro 2798: RHist mr0,mr;
2799: RHist r;
1.37 noro 2800: RHist *old_red;
1.71 noro 2801: ND_pairs s0,s,t;
1.43 noro 2802: EPOS oepos;
1.15 noro 2803:
1.1 noro 2804: obpe = nd_bpe;
1.11 noro 2805: oadv = nmv_adv;
1.43 noro 2806: oepos = nd_epos;
1.63 noro 2807: if ( obpe < 2 ) nd_bpe = 2;
2808: else if ( obpe < 3 ) nd_bpe = 3;
1.62 noro 2809: else if ( obpe < 4 ) nd_bpe = 4;
1.63 noro 2810: else if ( obpe < 5 ) nd_bpe = 5;
1.34 noro 2811: else if ( obpe < 6 ) nd_bpe = 6;
2812: else if ( obpe < 8 ) nd_bpe = 8;
1.63 noro 2813: else if ( obpe < 10 ) nd_bpe = 10;
1.34 noro 2814: else if ( obpe < 16 ) nd_bpe = 16;
2815: else if ( obpe < 32 ) nd_bpe = 32;
2816: else error("nd_reconstruct : exponent too large");
1.5 noro 2817:
1.73 noro 2818: nd_setup_parameters(nd_nvar,0);
1.1 noro 2819: prev_nm_free_list = _nm_free_list;
2820: prev_ndp_free_list = _ndp_free_list;
2821: _nm_free_list = 0;
2822: _ndp_free_list = 0;
1.53 noro 2823: for ( i = nd_psn-1; i >= 0; i-- ) ndv_realloc(nd_ps[i],obpe,oadv,oepos);
2824: if ( trace )
2825: for ( i = nd_psn-1; i >= 0; i-- )
2826: ndv_realloc(nd_ps_trace[i],obpe,oadv,oepos);
1.1 noro 2827: s0 = 0;
2828: for ( t = d; t; t = NEXT(t) ) {
2829: NEXTND_pairs(s0,s);
2830: s->i1 = t->i1;
2831: s->i2 = t->i2;
1.14 noro 2832: SG(s) = SG(t);
1.61 noro 2833: ndl_reconstruct(LCM(t),LCM(s),obpe,oepos);
1.1 noro 2834: }
1.37 noro 2835:
2836: old_red = (RHist *)ALLOCA(REDTAB_LEN*sizeof(RHist));
1.6 noro 2837: for ( i = 0; i < REDTAB_LEN; i++ ) {
1.37 noro 2838: old_red[i] = nd_red[i];
2839: nd_red[i] = 0;
2840: }
2841: for ( i = 0; i < REDTAB_LEN; i++ )
2842: for ( r = old_red[i]; r; r = NEXT(r) ) {
2843: NEWRHist(mr);
1.13 noro 2844: mr->index = r->index;
1.20 noro 2845: SG(mr) = SG(r);
1.61 noro 2846: ndl_reconstruct(DL(r),DL(mr),obpe,oepos);
1.37 noro 2847: h = ndl_hash_value(DL(mr));
2848: NEXT(mr) = nd_red[h];
2849: nd_red[h] = mr;
1.6 noro 2850: }
1.37 noro 2851: for ( i = 0; i < REDTAB_LEN; i++ ) old_red[i] = 0;
2852: old_red = 0;
1.11 noro 2853: for ( i = 0; i < nd_psn; i++ ) {
1.20 noro 2854: NEWRHist(r); SG(r) = SG(nd_psh[i]);
1.61 noro 2855: ndl_reconstruct(DL(nd_psh[i]),DL(r),obpe,oepos);
1.13 noro 2856: nd_psh[i] = r;
1.11 noro 2857: }
1.1 noro 2858: if ( s0 ) NEXT(s) = 0;
2859: prev_nm_free_list = 0;
2860: prev_ndp_free_list = 0;
1.71 noro 2861: #if 0
1.1 noro 2862: GC_gcollect();
1.71 noro 2863: #endif
1.1 noro 2864: return s0;
2865: }
2866:
1.61 noro 2867: void ndl_reconstruct(UINT *d,UINT *r,int obpe,EPOS oepos)
1.1 noro 2868: {
1.57 noro 2869: int n,i,ei,oepw,omask0,j,s,ord_l,l;
1.43 noro 2870: struct order_pair *op;
1.1 noro 2871:
2872: n = nd_nvar;
1.61 noro 2873: oepw = (sizeof(UINT)*8)/obpe;
1.43 noro 2874: omask0 = (1<<obpe)-1;
1.34 noro 2875: TD(r) = TD(d);
1.41 noro 2876: for ( i = nd_exporigin; i < nd_wpd; i++ ) r[i] = 0;
1.43 noro 2877: if ( nd_blockmask ) {
2878: l = nd_blockmask->n;
2879: op = nd_blockmask->order_pair;
2880: for ( i = 1; i < nd_exporigin; i++ )
2881: r[i] = d[i];
2882: for ( j = 0, s = 0; j < l; j++ ) {
2883: ord_l = op[j].length;
1.57 noro 2884: for ( i = 0; i < ord_l; i++, s++ ) {
2885: ei = GET_EXP_OLD(d,s);
2886: PUT_EXP(r,s,ei);
2887: }
1.1 noro 2888: }
1.43 noro 2889: } else {
1.56 noro 2890: for ( i = 0; i < n; i++ ) {
2891: ei = GET_EXP_OLD(d,i);
2892: PUT_EXP(r,i,ei);
2893: }
1.1 noro 2894: }
2895: }
1.3 noro 2896:
1.6 noro 2897: ND nd_copy(ND p)
2898: {
2899: NM m,mr,mr0;
1.41 noro 2900: int c,n;
1.6 noro 2901: ND r;
2902:
2903: if ( !p )
2904: return 0;
2905: else {
2906: for ( mr0 = 0, m = BDY(p); m; m = NEXT(m) ) {
2907: NEXTNM(mr0,mr);
1.14 noro 2908: CM(mr) = CM(m);
2909: ndl_copy(DL(m),DL(mr));
1.6 noro 2910: }
2911: NEXT(mr) = 0;
1.31 noro 2912: MKND(NV(p),mr0,LEN(p),r);
1.14 noro 2913: SG(r) = SG(p);
1.6 noro 2914: return r;
2915: }
2916: }
2917:
1.53 noro 2918: int nd_sp(int mod,int trace,ND_pairs p,ND *rp)
1.11 noro 2919: {
2920: NM m;
2921: NDV p1,p2;
2922: ND t1,t2;
1.61 noro 2923: UINT *lcm;
1.31 noro 2924: int td;
1.11 noro 2925:
1.53 noro 2926: if ( trace ) {
2927: p1 = nd_ps_trace[p->i1]; p2 = nd_ps_trace[p->i2];
2928: } else {
1.20 noro 2929: p1 = nd_ps[p->i1]; p2 = nd_ps[p->i2];
2930: }
1.34 noro 2931: lcm = LCM(p);
1.11 noro 2932: NEWNM(m);
1.20 noro 2933: CQ(m) = HCQ(p2);
1.34 noro 2934: ndl_sub(lcm,HDL(p1),DL(m));
1.56 noro 2935: if ( ndl_check_bound2(p->i1,DL(m)) )
2936: return 0;
1.55 noro 2937: t1 = ndv_mul_nm(mod,m,p1);
1.71 noro 2938: if ( mod == -1 ) CM(m) = _chsgnsf(HCM(p1));
2939: else if ( mod ) CM(m) = mod-HCM(p1);
1.34 noro 2940: else chsgnq(HCQ(p1),&CQ(m));
2941: ndl_sub(lcm,HDL(p2),DL(m));
1.14 noro 2942: if ( ndl_check_bound2(p->i2,DL(m)) ) {
1.11 noro 2943: nd_free(t1);
2944: return 0;
2945: }
1.55 noro 2946: t2 = ndv_mul_nm(mod,m,p2);
1.31 noro 2947: *rp = nd_add(mod,t1,t2);
1.11 noro 2948: FREENM(m);
2949: return 1;
2950: }
2951:
1.19 noro 2952: void ndv_mul_c(int mod,NDV p,int mul)
1.11 noro 2953: {
2954: NMV m;
2955: int c,c1,len,i;
2956:
1.34 noro 2957: if ( !p ) return;
1.14 noro 2958: len = LEN(p);
1.71 noro 2959: if ( mod == -1 )
2960: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) )
2961: CM(m) = _mulsf(CM(m),mul);
2962: else
2963: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2964: c1 = CM(m); DMAR(c1,mul,0,mod,c); CM(m) = c;
2965: }
1.11 noro 2966: }
2967:
1.16 noro 2968: void ndv_mul_c_q(NDV p,Q mul)
2969: {
2970: NMV m;
2971: Q c;
2972: int len,i;
2973:
1.34 noro 2974: if ( !p ) return;
1.16 noro 2975: len = LEN(p);
2976: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
2977: mulq(CQ(m),mul,&c); CQ(m) = c;
2978: }
2979: }
2980:
1.55 noro 2981: ND weyl_ndv_mul_nm(int mod,NM m0,NDV p) {
2982: int n2,i,j,l,n,tlen;
1.61 noro 2983: UINT *d0;
1.55 noro 2984: NM *tab,*psum;
2985: ND s,r;
2986: NM t;
2987: NMV m1;
2988:
2989: if ( !p ) return 0;
2990: n = NV(p); n2 = n>>1;
2991: d0 = DL(m0);
2992: l = LEN(p);
2993: for ( i = 0, tlen = 1; i < n2; i++ ) tlen *= (GET_EXP(d0,n2+i)+1);
2994: tab = (NM *)ALLOCA(tlen*sizeof(NM));
2995: psum = (NM *)ALLOCA(tlen*sizeof(NM));
2996: for ( i = 0; i < tlen; i++ ) psum[i] = 0;
1.56 noro 2997: m1 = (NMV)(((char *)BDY(p))+nmv_adv*(l-1));
2998: for ( i = l-1; i >= 0; i--, NMV_PREV(m1) ) {
1.55 noro 2999: /* m0(NM) * m1(NMV) => tab(NM) */
1.56 noro 3000: weyl_mul_nm_nmv(n,mod,m0,m1,tab,tlen);
1.55 noro 3001: for ( j = 0; j < tlen; j++ ) {
3002: if ( tab[j] ) {
3003: NEXT(tab[j]) = psum[j]; psum[j] = tab[j];
3004: }
3005: }
3006: }
3007: for ( i = tlen-1, r = 0; i >= 0; i-- )
3008: if ( psum[i] ) {
3009: for ( j = 0, t = psum[i]; t; t = NEXT(t), j++ );
3010: MKND(n,psum[i],j,s);
3011: r = nd_add(mod,r,s);
3012: }
1.56 noro 3013: if ( r ) SG(r) = SG(p)+TD(d0);
3014: return r;
1.55 noro 3015: }
3016:
1.56 noro 3017: /* product of monomials */
3018: /* XXX block order is not handled correctly */
3019:
1.55 noro 3020: void weyl_mul_nm_nmv(int n,int mod,NM m0,NMV m1,NM *tab,int tlen)
3021: {
1.56 noro 3022: int i,n2,j,s,curlen,homo,h,a,b,k,l,u,min;
1.61 noro 3023: UINT *d0,*d1,*d,*dt,*ctab;
1.56 noro 3024: Q *ctab_q;
3025: Q q,q1;
1.61 noro 3026: UINT c0,c1,c;
1.55 noro 3027: NM *p;
3028: NM m,t;
3029:
3030: for ( i = 0; i < tlen; i++ ) tab[i] = 0;
3031: if ( !m0 || !m1 ) return;
3032: d0 = DL(m0); d1 = DL(m1); n2 = n>>1;
3033: NEWNM(m); d = DL(m);
1.56 noro 3034: if ( mod ) {
3035: c0 = CM(m0); c1 = CM(m1); DMAR(c0,c1,0,mod,c); CM(m) = c;
3036: } else
3037: mulq(CQ(m0),CQ(m1),&CQ(m));
1.55 noro 3038: for ( i = 0; i < nd_wpd; i++ ) d[i] = 0;
3039: homo = n&1 ? 1 : 0;
3040: if ( homo ) {
3041: /* offset of h-degree */
3042: h = GET_EXP(d0,n-1)+GET_EXP(d1,n-1);
3043: PUT_EXP(DL(m),n-1,h);
3044: TD(DL(m)) = h;
1.61 noro 3045: if ( nd_blockmask ) ndl_weight_mask(DL(m));
1.55 noro 3046: }
3047: tab[0] = m;
3048: NEWNM(m); d = DL(m);
1.57 noro 3049: for ( i = 0, curlen = 1; i < n2; i++ ) {
1.55 noro 3050: a = GET_EXP(d0,i); b = GET_EXP(d1,n2+i);
3051: k = GET_EXP(d0,n2+i); l = GET_EXP(d1,i);
3052: /* xi^a*(Di^k*xi^l)*Di^b */
3053: a += l; b += k;
1.56 noro 3054: s = MUL_WEIGHT(a,i)+MUL_WEIGHT(b,n2+i);
1.55 noro 3055: if ( !k || !l ) {
3056: for ( j = 0; j < curlen; j++ )
1.56 noro 3057: if ( t = tab[j] ) {
3058: dt = DL(t);
3059: PUT_EXP(dt,i,a); PUT_EXP(dt,n2+i,b); TD(dt) += s;
1.61 noro 3060: if ( nd_blockmask ) ndl_weight_mask(dt);
1.55 noro 3061: }
3062: curlen *= k+1;
3063: continue;
3064: }
3065: min = MIN(k,l);
1.56 noro 3066: if ( mod ) {
1.61 noro 3067: ctab = (UINT *)ALLOCA((min+1)*sizeof(UINT));
1.56 noro 3068: mkwcm(k,l,mod,ctab);
3069: } else {
3070: ctab_q = (Q *)ALLOCA((min+1)*sizeof(Q));
3071: mkwc(k,l,ctab_q);
3072: }
1.57 noro 3073: for ( j = min; j >= 0; j-- ) {
1.56 noro 3074: for ( u = 0; u < nd_wpd; u++ ) d[u] = 0;
1.55 noro 3075: PUT_EXP(d,i,a-j); PUT_EXP(d,n2+i,b-j);
1.56 noro 3076: h = MUL_WEIGHT(a-j,i)+MUL_WEIGHT(b-j,n2+i);
1.55 noro 3077: if ( homo ) {
3078: TD(d) = s;
1.56 noro 3079: PUT_EXP(d,n-1,s-h);
1.55 noro 3080: } else TD(d) = h;
1.61 noro 3081: if ( nd_blockmask ) ndl_weight_mask(d);
1.56 noro 3082: if ( mod ) c = ctab[j];
3083: else q = ctab_q[j];
1.57 noro 3084: p = tab+curlen*j;
3085: if ( j == 0 ) {
3086: for ( u = 0; u < curlen; u++, p++ ) {
3087: if ( tab[u] ) {
3088: ndl_addto(DL(tab[u]),d);
3089: if ( mod ) {
3090: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(tab[u]) = c1;
3091: } else {
3092: mulq(CQ(tab[u]),q,&q1); CQ(tab[u]) = q1;
3093: }
3094: }
1.56 noro 3095: }
1.57 noro 3096: } else {
3097: for ( u = 0; u < curlen; u++, p++ ) {
3098: if ( tab[u] ) {
3099: NEWNM(t);
3100: ndl_add(DL(tab[u]),d,DL(t));
3101: if ( mod ) {
3102: c0 = CM(tab[u]); DMAR(c0,c,0,mod,c1); CM(t) = c1;
3103: } else
3104: mulq(CQ(tab[u]),q,&CQ(t));
3105: *p = t;
3106: }
1.55 noro 3107: }
3108: }
3109: }
3110: curlen *= k+1;
3111: }
3112: FREENM(m);
3113: }
3114:
1.63 noro 3115: ND ndv_mul_nm_symbolic(NM m0,NDV p)
3116: {
3117: NM mr,mr0;
3118: NMV m;
3119: UINT *d,*dt,*dm;
3120: int c,n,td,i,c1,c2,len;
3121: Q q;
3122: ND r;
3123:
3124: if ( !p ) return 0;
3125: else {
3126: n = NV(p); m = BDY(p);
3127: d = DL(m0);
3128: len = LEN(p);
3129: mr0 = 0;
3130: td = TD(d);
3131: c = CM(m0);
3132: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3133: NEXTNM(mr0,mr);
3134: CM(mr) = 1;
3135: ndl_add(DL(m),d,DL(mr));
3136: }
3137: NEXT(mr) = 0;
3138: MKND(NV(p),mr0,len,r);
3139: SG(r) = SG(p) + TD(d);
3140: return r;
3141: }
3142: }
3143:
1.55 noro 3144: ND ndv_mul_nm(int mod,NM m0,NDV p)
1.9 noro 3145: {
3146: NM mr,mr0;
3147: NMV m;
1.61 noro 3148: UINT *d,*dt,*dm;
1.9 noro 3149: int c,n,td,i,c1,c2,len;
1.16 noro 3150: Q q;
1.9 noro 3151: ND r;
3152:
1.34 noro 3153: if ( !p ) return 0;
1.55 noro 3154: else if ( do_weyl )
1.71 noro 3155: if ( mod == -1 )
3156: error("ndv_mul_nm : not implemented (weyl)");
3157: else
3158: return weyl_ndv_mul_nm(mod,m0,p);
1.9 noro 3159: else {
3160: n = NV(p); m = BDY(p);
1.34 noro 3161: d = DL(m0);
1.14 noro 3162: len = LEN(p);
1.9 noro 3163: mr0 = 0;
1.34 noro 3164: td = TD(d);
1.71 noro 3165: if ( mod == -1 ) {
3166: c = CM(m0);
3167: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3168: NEXTNM(mr0,mr);
3169: CM(mr) = _mulsf(CM(m),c);
3170: ndl_add(DL(m),d,DL(mr));
3171: }
3172: } else if ( mod ) {
1.16 noro 3173: c = CM(m0);
3174: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3175: NEXTNM(mr0,mr);
3176: c1 = CM(m);
1.19 noro 3177: DMAR(c1,c,0,mod,c2);
1.16 noro 3178: CM(mr) = c2;
3179: ndl_add(DL(m),d,DL(mr));
3180: }
3181: } else {
3182: q = CQ(m0);
3183: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
3184: NEXTNM(mr0,mr);
3185: mulq(CQ(m),q,&CQ(mr));
3186: ndl_add(DL(m),d,DL(mr));
3187: }
1.4 noro 3188: }
1.9 noro 3189: NEXT(mr) = 0;
1.31 noro 3190: MKND(NV(p),mr0,len,r);
1.34 noro 3191: SG(r) = SG(p) + TD(d);
1.9 noro 3192: return r;
1.4 noro 3193: }
3194: }
3195:
1.43 noro 3196: void ndv_realloc(NDV p,int obpe,int oadv,EPOS oepos)
1.11 noro 3197: {
1.13 noro 3198: NMV m,mr,mr0,t;
3199: int len,i,k;
1.11 noro 3200:
1.61 noro 3201: if ( !p ) return;
3202: m = BDY(p); len = LEN(p);
3203: mr0 = nmv_adv>oadv?(NMV)REALLOC(BDY(p),len*nmv_adv):BDY(p);
3204: m = (NMV)((char *)mr0+(len-1)*oadv);
3205: mr = (NMV)((char *)mr0+(len-1)*nmv_adv);
3206: t = (NMV)ALLOCA(nmv_adv);
3207: for ( i = 0; i < len; i++, NMV_OPREV(m), NMV_PREV(mr) ) {
3208: CQ(t) = CQ(m);
3209: for ( k = 0; k < nd_wpd; k++ ) DL(t)[k] = 0;
3210: ndl_reconstruct(DL(m),DL(t),obpe,oepos);
3211: CQ(mr) = CQ(t);
3212: ndl_copy(DL(t),DL(mr));
3213: }
3214: BDY(p) = mr0;
3215: }
3216:
3217: NDV ndv_dup_realloc(NDV p,int obpe,int oadv,EPOS oepos)
3218: {
3219: NMV m,mr,mr0;
3220: int len,i;
3221: NDV r;
1.11 noro 3222:
1.61 noro 3223: if ( !p ) return 0;
3224: m = BDY(p); len = LEN(p);
3225: mr0 = mr = (NMV)MALLOC(len*nmv_adv);
3226: for ( i = 0; i < len; i++, NMV_OADV(m), NMV_ADV(mr) ) {
3227: ndl_zero(DL(mr));
3228: ndl_reconstruct(DL(m),DL(mr),obpe,oepos);
3229: CQ(mr) = CQ(m);
1.11 noro 3230: }
1.61 noro 3231: MKNDV(NV(p),mr0,len,r);
3232: SG(r) = SG(p);
3233: return r;
1.11 noro 3234: }
3235:
1.61 noro 3236: /* duplicate p */
3237:
3238: NDV ndv_dup(int mod,NDV p)
1.3 noro 3239: {
3240: NDV d;
1.61 noro 3241: NMV t,m,m0;
1.3 noro 3242: int i,len;
3243:
1.34 noro 3244: if ( !p ) return 0;
1.31 noro 3245: len = LEN(p);
1.34 noro 3246: m0 = m = (NMV)(mod?MALLOC_ATOMIC(len*nmv_adv):MALLOC(len*nmv_adv));
1.61 noro 3247: for ( t = BDY(p), i = 0; i < len; i++, NMV_ADV(t), NMV_ADV(m) ) {
1.14 noro 3248: ndl_copy(DL(t),DL(m));
1.16 noro 3249: CQ(m) = CQ(t);
1.3 noro 3250: }
3251: MKNDV(NV(p),m0,len,d);
1.23 noro 3252: SG(d) = SG(p);
3253: return d;
3254: }
3255:
1.63 noro 3256: ND nd_dup(ND p)
3257: {
3258: ND d;
3259: NM t,m,m0;
3260:
3261: if ( !p ) return 0;
3262: for ( m0 = 0, t = BDY(p); t; t = NEXT(t) ) {
3263: NEXTNM(m0,m);
3264: ndl_copy(DL(t),DL(m));
3265: CQ(m) = CQ(t);
3266: }
3267: if ( m0 ) NEXT(m) = 0;
3268: MKND(NV(p),m0,LEN(p),d);
3269: SG(d) = SG(p);
3270: return d;
3271: }
3272:
1.61 noro 3273: /* XXX if p->len == 0 then it represents 0 */
3274:
3275: void ndv_mod(int mod,NDV p)
3276: {
3277: NMV t,d;
3278: int r;
3279: int i,len,dlen;
1.71 noro 3280: Obj gfs;
1.61 noro 3281:
3282: if ( !p ) return;
3283: len = LEN(p);
3284: dlen = 0;
1.71 noro 3285: if ( mod == -1 )
3286: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3287: simp_ff((Obj)CP(t),&gfs);
3288: r = FTOIF(CONT((GFS)gfs));
1.61 noro 3289: CM(d) = r;
3290: ndl_copy(DL(t),DL(d));
3291: NMV_ADV(d);
3292: dlen++;
3293: }
1.71 noro 3294: else
3295: for ( t = d = BDY(p), i = 0; i < len; i++, NMV_ADV(t) ) {
3296: r = rem(NM(CQ(t)),mod);
3297: if ( r ) {
3298: if ( SGN(CQ(t)) < 0 )
3299: r = mod-r;
3300: CM(d) = r;
3301: ndl_copy(DL(t),DL(d));
3302: NMV_ADV(d);
3303: dlen++;
3304: }
3305: }
1.61 noro 3306: LEN(p) = dlen;
3307: }
3308:
3309: NDV ptondv(VL vl,VL dvl,P p)
3310: {
3311: ND nd;
3312:
3313: nd = ptond(vl,dvl,p);
3314: return ndtondv(0,nd);
3315: }
3316:
3317: ND ptond(VL vl,VL dvl,P p)
1.23 noro 3318: {
1.61 noro 3319: int n,i,j,k,e;
3320: VL tvl;
3321: V v;
3322: DCP dc;
3323: DCP *w;
3324: ND r,s,t,u;
3325: P x;
3326: int c;
3327: UINT *d;
1.23 noro 3328: NM m,m0;
1.61 noro 3329:
3330: if ( !p )
3331: return 0;
3332: else if ( NUM(p) ) {
3333: NEWNM(m);
3334: ndl_zero(DL(m));
3335: CQ(m) = (Q)p;
3336: NEXT(m) = 0;
3337: MKND(nd_nvar,m,1,r);
3338: SG(r) = 0;
3339: return r;
3340: } else {
3341: for ( dc = DC(p), k = 0; dc; dc = NEXT(dc), k++ );
3342: w = (DCP *)ALLOCA(k*sizeof(DCP));
3343: for ( dc = DC(p), j = 0; j < k; dc = NEXT(dc), j++ ) w[j] = dc;
3344: for ( i = 0, tvl = dvl, v = VR(p);
3345: vl && tvl->v != v; tvl = NEXT(tvl), i++ );
3346: if ( !tvl ) {
3347: for ( j = k-1, s = 0, MKV(v,x); j >= 0; j-- ) {
3348: t = ptond(vl,dvl,COEF(w[j]));
3349: pwrp(vl,x,DEG(w[j]),&p);
3350: nd_mul_c_p(CO,t,p); s = nd_add(0,s,t);
3351: }
3352: return s;
3353: } else {
3354: NEWNM(m0); d = DL(m0);
3355: for ( j = k-1, s = 0; j >= 0; j-- ) {
3356: ndl_zero(d); e = QTOS(DEG(w[j])); PUT_EXP(d,i,e);
3357: TD(d) = MUL_WEIGHT(e,i);
3358: if ( nd_blockmask) ndl_weight_mask(d);
3359: t = ptond(vl,dvl,COEF(w[j]));
3360: for ( m = BDY(t); m; m = NEXT(m) )
3361: ndl_addto(DL(m),d);
3362: SG(t) += TD(d);
3363: s = nd_add(0,s,t);
3364: }
3365: FREENM(m0);
3366: return s;
3367: }
3368: }
3369: }
3370:
3371: P ndvtop(int mod,VL vl,VL dvl,NDV p)
3372: {
3373: VL tvl;
3374: int len,n,j,i,e;
3375: NMV m;
3376: Q q;
3377: P c;
3378: UINT *d;
3379: P s,r,u,t,w;
1.71 noro 3380: GFS gfs;
1.23 noro 3381:
1.34 noro 3382: if ( !p ) return 0;
1.61 noro 3383: else {
3384: len = LEN(p);
3385: n = NV(p);
3386: m = (NMV)(((char *)BDY(p))+nmv_adv*(len-1));
3387: for ( j = len-1, s = 0; j >= 0; j--, NMV_PREV(m) ) {
1.71 noro 3388: if ( mod == -1 ) {
3389: e = IFTOF(CM(m)); MKGFS(e,gfs); c = (P)gfs;
3390: } else if ( mod ) {
1.61 noro 3391: STOQ(CM(m),q); c = (P)q;
3392: } else
3393: c = CP(m);
3394: d = DL(m);
3395: for ( i = 0, t = c, tvl = dvl; i < n; tvl = NEXT(tvl), i++ ) {
3396: MKV(tvl->v,r); e = GET_EXP(d,i); STOQ(e,q);
3397: pwrp(vl,r,q,&u); mulp(vl,t,u,&w); t = w;
3398: }
3399: addp(vl,s,t,&u); s = u;
3400: }
3401: return s;
1.23 noro 3402: }
1.3 noro 3403: }
3404:
1.61 noro 3405: NDV ndtondv(int mod,ND p)
1.11 noro 3406: {
3407: NDV d;
1.61 noro 3408: NMV m,m0;
3409: NM t;
3410: int i,len;
1.11 noro 3411:
1.34 noro 3412: if ( !p ) return 0;
1.61 noro 3413: len = LEN(p);
1.74 ! noro 3414: if ( mod )
! 3415: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(len*nmv_adv);
! 3416: else
! 3417: m0 = m = MALLOC(len*nmv_adv);
! 3418: ndv_alloc += nmv_adv*len;
1.61 noro 3419: for ( t = BDY(p), i = 0; t; t = NEXT(t), i++, NMV_ADV(m) ) {
3420: ndl_copy(DL(t),DL(m));
3421: CQ(m) = CQ(t);
1.11 noro 3422: }
1.61 noro 3423: MKNDV(NV(p),m0,len,d);
1.14 noro 3424: SG(d) = SG(p);
1.11 noro 3425: return d;
3426: }
3427:
1.61 noro 3428: ND ndvtond(int mod,NDV p)
1.11 noro 3429: {
1.61 noro 3430: ND d;
3431: NM m,m0;
1.11 noro 3432: NMV t;
1.61 noro 3433: int i,len;
1.11 noro 3434:
1.34 noro 3435: if ( !p ) return 0;
1.11 noro 3436: m0 = 0;
1.61 noro 3437: len = p->len;
3438: for ( t = BDY(p), i = 0; i < len; NMV_ADV(t), i++ ) {
3439: NEXTNM(m0,m);
3440: ndl_copy(DL(t),DL(m));
3441: CQ(m) = CQ(t);
1.11 noro 3442: }
3443: NEXT(m) = 0;
1.61 noro 3444: MKND(NV(p),m0,len,d);
1.14 noro 3445: SG(d) = SG(p);
1.11 noro 3446: return d;
3447: }
3448:
1.3 noro 3449: void ndv_print(NDV p)
3450: {
3451: NMV m;
3452: int i,len;
3453:
1.34 noro 3454: if ( !p ) printf("0\n");
1.3 noro 3455: else {
1.14 noro 3456: len = LEN(p);
1.3 noro 3457: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
1.71 noro 3458: if ( CM(m) & 0x80000000 ) printf("+@_%d*",IFTOF(CM(m)));
3459: else printf("+%d*",CM(m));
1.16 noro 3460: ndl_print(DL(m));
3461: }
3462: printf("\n");
3463: }
3464: }
3465:
3466: void ndv_print_q(NDV p)
3467: {
3468: NMV m;
3469: int i,len;
3470:
1.34 noro 3471: if ( !p ) printf("0\n");
1.16 noro 3472: else {
3473: len = LEN(p);
3474: for ( m = BDY(p), i = 0; i < len; i++, NMV_ADV(m) ) {
3475: printf("+");
1.74 ! noro 3476: printexpr(CO,(Obj)CQ(m));
1.16 noro 3477: printf("*");
1.14 noro 3478: ndl_print(DL(m));
1.3 noro 3479: }
3480: printf("\n");
3481: }
1.25 noro 3482: }
3483:
1.61 noro 3484: NODE ndv_reducebase(NODE x)
1.27 noro 3485: {
3486: int len,i,j;
3487: NDV *w;
3488: NODE t,t0;
3489:
3490: len = length(x);
3491: w = (NDV *)ALLOCA(len*sizeof(NDV));
3492: for ( i = 0, t = x; i < len; i++, t = NEXT(t) ) w[i] = BDY(t);
3493: for ( i = 0; i < len; i++ ) {
3494: for ( j = 0; j < i; j++ ) {
3495: if ( w[i] && w[j] )
3496: if ( ndl_reducible(HDL(w[i]),HDL(w[j])) ) w[i] = 0;
3497: else if ( ndl_reducible(HDL(w[j]),HDL(w[i])) ) w[j] = 0;
3498: }
3499: }
3500: for ( i = len-1, t0 = 0; i >= 0; i-- ) {
3501: if ( w[i] ) { NEXTNODE(t0,t); BDY(t) = (pointer)w[i]; }
3502: }
3503: NEXT(t) = 0; x = t0;
3504: return x;
1.11 noro 3505: }
1.32 noro 3506:
1.43 noro 3507: /* XXX incomplete */
3508:
1.32 noro 3509: void nd_init_ord(struct order_spec *ord)
3510: {
1.43 noro 3511: switch ( ord->id ) {
1.32 noro 3512: case 0:
1.43 noro 3513: switch ( ord->ord.simple ) {
3514: case 0:
3515: nd_dcomp = 1;
3516: nd_isrlex = 1;
3517: break;
3518: case 1:
3519: nd_dcomp = 1;
3520: nd_isrlex = 0;
3521: break;
3522: case 2:
3523: nd_dcomp = 0;
3524: nd_isrlex = 0;
1.45 noro 3525: ndl_compare_function = ndl_lex_compare;
1.58 noro 3526: break;
3527: case 11:
3528: /* XXX */
3529: nd_dcomp = 0;
3530: nd_isrlex = 1;
3531: ndl_compare_function = ndl_ww_lex_compare;
1.43 noro 3532: break;
3533: default:
3534: error("nd_gr : unsupported order");
3535: }
1.32 noro 3536: break;
3537: case 1:
1.43 noro 3538: /* XXX */
3539: nd_dcomp = -1;
1.32 noro 3540: nd_isrlex = 0;
1.45 noro 3541: ndl_compare_function = ndl_block_compare;
1.34 noro 3542: break;
1.43 noro 3543: case 2:
3544: error("nd_init_ord : matrix order is not supported yet.");
1.32 noro 3545: break;
3546: }
1.41 noro 3547: nd_ord = ord;
1.32 noro 3548: }
3549:
1.43 noro 3550: BlockMask nd_create_blockmask(struct order_spec *ord)
3551: {
3552: int n,i,j,s,l;
1.61 noro 3553: UINT *t;
1.43 noro 3554: BlockMask bm;
3555:
3556: if ( !ord->id )
3557: return 0;
3558: n = ord->ord.block.length;
3559: bm = (BlockMask)MALLOC(sizeof(struct oBlockMask));
3560: bm->n = n;
3561: bm->order_pair = ord->ord.block.order_pair;
1.61 noro 3562: bm->mask = (UINT **)MALLOC(n*sizeof(UINT *));
1.43 noro 3563: for ( i = 0, s = 0; i < n; i++ ) {
1.61 noro 3564: bm->mask[i] = t = (UINT *)MALLOC_ATOMIC(nd_wpd*sizeof(UINT));
1.43 noro 3565: for ( j = 0; j < nd_wpd; j++ ) t[j] = 0;
3566: l = bm->order_pair[i].length;
3567: for ( j = 0; j < l; j++, s++ ) PUT_EXP(t,s,nd_mask0);
3568: }
3569: return bm;
1.57 noro 3570: }
3571:
3572: EPOS nd_create_epos(struct order_spec *ord)
3573: {
3574: int i,j,l,s,ord_l,ord_o;
3575: EPOS epos;
3576: struct order_pair *op;
3577:
3578: epos = (EPOS)MALLOC_ATOMIC(nd_nvar*sizeof(struct oEPOS));
3579: switch ( ord->id ) {
3580: case 0:
3581: if ( nd_isrlex ) {
3582: for ( i = 0; i < nd_nvar; i++ ) {
3583: epos[i].i = nd_exporigin + (nd_nvar-1-i)/nd_epw;
3584: epos[i].s = (nd_epw-((nd_nvar-1-i)%nd_epw)-1)*nd_bpe;
3585: }
3586: } else {
3587: for ( i = 0; i < nd_nvar; i++ ) {
3588: epos[i].i = nd_exporigin + i/nd_epw;
3589: epos[i].s = (nd_epw-(i%nd_epw)-1)*nd_bpe;
3590: }
3591: }
3592: break;
3593: case 1:
3594: /* block order */
3595: l = ord->ord.block.length;
3596: op = ord->ord.block.order_pair;
3597: for ( j = 0, s = 0; j < l; j++ ) {
3598: ord_o = op[j].order;
3599: ord_l = op[j].length;
3600: if ( !ord_o )
3601: for ( i = 0; i < ord_l; i++ ) {
3602: epos[s+i].i = nd_exporigin + (s+ord_l-i-1)/nd_epw;
3603: epos[s+i].s = (nd_epw-((s+ord_l-i-1)%nd_epw)-1)*nd_bpe;
3604: }
3605: else
3606: for ( i = 0; i < ord_l; i++ ) {
3607: epos[s+i].i = nd_exporigin + (s+i)/nd_epw;
3608: epos[s+i].s = (nd_epw-((s+i)%nd_epw)-1)*nd_bpe;
3609: }
3610: s += ord_l;
3611: }
3612: break;
3613: case 2:
3614: error("nd_create_epos : matrix order is not supported yet.");
3615: }
3616: return epos;
1.43 noro 3617: }
1.59 noro 3618:
3619: /* external interface */
3620:
3621: void nd_nf_p(P f,LIST g,LIST v,int m,struct order_spec *ord,P *rp)
3622: {
1.61 noro 3623: NODE t,in0,in;
1.59 noro 3624: ND nd,nf;
1.61 noro 3625: NDV ndv;
3626: VL vv,tv;
3627: int stat,nvar,max,e;
1.69 noro 3628: union oNDC dn;
1.59 noro 3629:
3630: pltovl(v,&vv);
1.74 ! noro 3631: for ( nvar = 0, tv = vv; tv; tv = NEXT(tv), nvar++ );
1.61 noro 3632:
3633: /* get the degree bound */
3634: for ( t = BDY(g), max = 0; t; t = NEXT(t) )
3635: for ( tv = vv; tv; tv = NEXT(tv) ) {
3636: e = getdeg(tv->v,(P)BDY(t));
3637: max = MAX(e,max);
3638: }
3639: for ( tv = vv; tv; tv = NEXT(tv) ) {
3640: e = getdeg(tv->v,f);
3641: max = MAX(e,max);
3642: }
3643:
1.59 noro 3644: nd_init_ord(ord);
1.61 noro 3645: nd_setup_parameters(nvar,max);
3646:
3647: /* conversion to ndv */
3648: for ( in0 = 0, t = BDY(g); t; t = NEXT(t) ) {
3649: NEXTNODE(in0,in);
3650: BDY(in) = (pointer)ptondv(CO,vv,(P)BDY(t));
1.69 noro 3651: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3652: }
3653: NEXTNODE(in0,in);
3654: BDY(in) = (pointer)ptondv(CO,vv,f);
1.69 noro 3655: if ( m ) ndv_mod(m,(NDV)BDY(in));
1.61 noro 3656: NEXT(in) = 0;
3657:
3658: ndv_setup(m,0,in0);
1.59 noro 3659: nd_psn--;
3660: nd_scale=2;
3661: while ( 1 ) {
3662: nd = (pointer)ndvtond(m,nd_ps[nd_psn]);
1.69 noro 3663: stat = nd_nf(m,nd,nd_ps,1,0,&nf);
1.59 noro 3664: if ( !stat ) {
3665: nd_psn++;
3666: nd_reconstruct(m,0,0);
3667: nd_psn--;
3668: } else
3669: break;
3670: }
1.61 noro 3671: *rp = ndvtop(m,CO,vv,ndtondv(m,nf));
1.63 noro 3672: }
3673:
3674: int nd_to_vect(int mod,UINT *s0,int n,ND d,UINT *r)
3675: {
3676: NM m;
3677: UINT *t,*s;
3678: int i;
3679:
3680: for ( i = 0; i < n; i++ ) r[i] = 0;
3681: for ( i = 0, s = s0, m = BDY(d); m; m = NEXT(m) ) {
3682: t = DL(m);
3683: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3684: r[i] = CM(m);
3685: }
3686: for ( i = 0; !r[i]; i++ );
3687: return i;
3688: }
3689:
1.74 ! noro 3690: int ndv_to_vect(int mod,UINT *s0,int n,NDV d,UINT *r)
! 3691: {
! 3692: NMV m;
! 3693: UINT *t,*s;
! 3694: int i,j,len;
! 3695:
! 3696: for ( i = 0; i < n; i++ ) r[i] = 0;
! 3697: m = BDY(d);
! 3698: len = LEN(d);
! 3699: for ( i = j = 0, s = s0; j < len; j++, NMV_ADV(m)) {
! 3700: t = DL(m);
! 3701: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
! 3702: r[i] = CM(m);
! 3703: }
! 3704: for ( i = 0; !r[i]; i++ );
! 3705: return i;
! 3706: }
! 3707:
1.63 noro 3708: int nm_ind_pair_to_vect(int mod,UINT *s0,int n,NM_ind_pair pair,UINT *r)
3709: {
3710: NM m;
3711: NMV mr;
3712: UINT *d,*t,*s;
3713: NDV p;
3714: int i,j,len;
3715:
3716: m = pair->mul;
3717: d = DL(m);
3718: p = nd_ps[pair->index];
3719: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
3720: for ( i = 0; i < n; i++ ) r[i] = 0;
3721: len = LEN(p);
3722: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3723: ndl_add(d,DL(mr),t);
3724: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
3725: r[i] = CM(mr);
3726: }
3727: for ( i = 0; !r[i]; i++ );
3728: return i;
3729: }
3730:
1.67 noro 3731: IndArray nm_ind_pair_to_vect_compress(int mod,UINT *s0,int n,NM_ind_pair pair)
1.64 noro 3732: {
3733: NM m;
3734: NMV mr;
3735: UINT *d,*t,*s;
3736: NDV p;
1.67 noro 3737: unsigned char *ivc;
3738: unsigned short *ivs;
3739: UINT *v,*ivi;
3740: int i,j,len,prev,diff,cdiff;
3741: IndArray r;
1.64 noro 3742:
3743: m = pair->mul;
3744: d = DL(m);
3745: p = nd_ps[pair->index];
3746: len = LEN(p);
3747: t = (UINT *)ALLOCA(nd_wpd*sizeof(UINT));
1.67 noro 3748: r = (IndArray)MALLOC(sizeof(struct oIndArray));
3749: v = (unsigned int *)ALLOCA(len*sizeof(unsigned int));
1.64 noro 3750: for ( i = j = 0, s = s0, mr = BDY(p); j < len; j++, NMV_ADV(mr) ) {
3751: ndl_add(d,DL(mr),t);
3752: for ( ; !ndl_equal(t,s); s += nd_wpd, i++ );
1.67 noro 3753: v[j] = i;
3754: }
3755: r->head = v[0];
3756: diff = 0;
3757: for ( i = 1; i < len; i++ ) {
3758: cdiff = v[i]-v[i-1]; diff = MAX(cdiff,diff);
3759: }
3760: if ( diff < 256 ) {
3761: r->width = 1;
3762: ivc = (unsigned char *)MALLOC_ATOMIC(len*sizeof(unsigned char));
3763: r->index.c = ivc;
3764: for ( i = 1, ivc[0] = 0; i < len; i++ ) ivc[i] = v[i]-v[i-1];
3765: } else if ( diff < 65536 ) {
3766: r->width = 2;
3767: ivs = (unsigned short *)MALLOC_ATOMIC(len*sizeof(unsigned short));
3768: r->index.s = ivs;
3769: for ( i = 1, ivs[0] = 0; i < len; i++ ) ivs[i] = v[i]-v[i-1];
3770: } else {
3771: r->width = 4;
3772: ivi = (unsigned int *)MALLOC_ATOMIC(len*sizeof(unsigned int));
3773: r->index.i = ivi;
3774: for ( i = 1, ivi[0] = 0; i < len; i++ ) ivi[i] = v[i]-v[i-1];
1.64 noro 3775: }
1.67 noro 3776: return r;
1.64 noro 3777: }
3778:
1.65 noro 3779:
1.74 ! noro 3780: void ndv_reduce_vect(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.65 noro 3781: {
1.67 noro 3782: int i,j,k,len,pos,prev;
1.66 noro 3783: UINT c,c1,c2,c3,up,lo,dmy;
1.67 noro 3784: IndArray ivect;
3785: unsigned char *ivc;
3786: unsigned short *ivs;
3787: unsigned int *ivi;
1.65 noro 3788: NDV redv;
1.67 noro 3789: NMV mr;
1.65 noro 3790: NODE rp;
3791:
1.74 ! noro 3792: for ( i = 0; i < nred; i++ ) {
1.65 noro 3793: ivect = imat[i];
1.67 noro 3794: k = ivect->head; svect[k] %= m;
1.65 noro 3795: if ( c = svect[k] ) {
1.74 ! noro 3796: c = m-c; redv = nd_ps[rp0[i]->index];
1.67 noro 3797: len = LEN(redv); mr = BDY(redv);
3798: svect[k] = 0; prev = k;
3799: switch ( ivect->width ) {
3800: case 1:
3801: ivc = ivect->index.c;
3802: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3803: pos = prev+ivc[j]; c1 = CM(mr); c2 = svect[pos];
3804: prev = pos;
3805: DMA(c1,c,c2,up,lo);
3806: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3807: } else svect[pos] = lo;
3808: }
3809: break;
3810: case 2:
3811: ivs = ivect->index.s;
3812: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3813: pos = prev+ivs[j]; c1 = CM(mr); c2 = svect[pos];
3814: prev = pos;
3815: DMA(c1,c,c2,up,lo);
3816: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3817: } else svect[pos] = lo;
3818: }
3819: break;
3820: case 4:
3821: ivi = ivect->index.i;
3822: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3823: pos = prev+ivi[j]; c1 = CM(mr); c2 = svect[pos];
3824: prev = pos;
3825: DMA(c1,c,c2,up,lo);
3826: if ( up ) { DSAB(m,up,lo,dmy,c3); svect[pos] = c3;
3827: } else svect[pos] = lo;
3828: }
3829: break;
1.65 noro 3830: }
3831: }
3832: }
1.66 noro 3833: for ( i = 0; i < col; i++ )
3834: if ( svect[i] >= (UINT)m ) svect[i] %= m;
1.65 noro 3835: }
3836:
1.74 ! noro 3837: void ndv_reduce_vect_sf(int m,UINT *svect,int col,IndArray *imat,NM_ind_pair *rp0,int nred)
1.72 noro 3838: {
3839: int i,j,k,len,pos,prev;
3840: UINT c,c1,c2,c3,up,lo,dmy;
3841: IndArray ivect;
3842: unsigned char *ivc;
3843: unsigned short *ivs;
3844: unsigned int *ivi;
3845: NDV redv;
3846: NMV mr;
3847: NODE rp;
3848:
1.74 ! noro 3849: for ( i = 0; i < nred; i++ ) {
1.72 noro 3850: ivect = imat[i];
3851: k = ivect->head; svect[k] %= m;
3852: if ( c = svect[k] ) {
1.74 ! noro 3853: c = _chsgnsf(c); redv = nd_ps[rp0[i]->index];
1.72 noro 3854: len = LEN(redv); mr = BDY(redv);
3855: svect[k] = 0; prev = k;
3856: switch ( ivect->width ) {
3857: case 1:
3858: ivc = ivect->index.c;
3859: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3860: pos = prev+ivc[j]; prev = pos;
3861: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3862: }
3863: break;
3864: case 2:
3865: ivs = ivect->index.s;
3866: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3867: pos = prev+ivs[j]; prev = pos;
3868: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3869: }
3870: break;
3871: case 4:
3872: ivi = ivect->index.i;
3873: for ( j = 1, NMV_ADV(mr); j < len; j++, NMV_ADV(mr) ) {
3874: pos = prev+ivi[j]; prev = pos;
3875: svect[pos] = _addsf(_mulsf(CM(mr),c),svect[pos]);
3876: }
3877: break;
3878: }
3879: }
3880: }
3881: }
3882:
1.65 noro 3883: NDV vect_to_ndv(UINT *vect,int spcol,int col,int *rhead,UINT *s0vect)
3884: {
3885: int j,k,len;
3886: UINT *p;
3887: UINT c;
3888: NDV r;
3889: NMV mr0,mr;
3890:
3891: for ( j = 0, len = 0; j < spcol; j++ ) if ( vect[j] ) len++;
3892: if ( !len ) return 0;
3893: else {
1.74 ! noro 3894: mr0 = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
! 3895: ndv_alloc += nmv_adv*len;
1.65 noro 3896: mr = mr0;
3897: p = s0vect;
3898: for ( j = k = 0; j < col; j++, p += nd_wpd )
3899: if ( !rhead[j] ) {
3900: if ( c = vect[k++] ) {
3901: ndl_copy(p,DL(mr)); CM(mr) = c; NMV_ADV(mr);
3902: }
3903: }
3904: MKNDV(nd_nvar,mr0,len,r);
3905: return r;
3906: }
3907: }
3908:
1.74 ! noro 3909: int nd_sp_f4(int m,ND_pairs l,PGeoBucket bucket)
1.65 noro 3910: {
3911: ND_pairs t;
3912: NODE sp0,sp;
3913: int stat;
3914: ND spol;
3915:
3916: for ( t = l; t; t = NEXT(t) ) {
3917: stat = nd_sp(m,0,t,&spol);
3918: if ( !stat ) return 0;
3919: if ( spol ) {
3920: add_pbucket_symbolic(bucket,spol);
3921: }
3922: }
1.68 noro 3923: return 1;
1.65 noro 3924: }
3925:
3926: int nd_symbolic_preproc(PGeoBucket bucket,UINT **s0vect,NODE *r)
3927: {
3928: NODE rp0,rp;
3929: NM mul,head,s0,s;
3930: int index,col,i;
3931: RHist h;
3932: UINT *s0v,*p;
3933: NM_ind_pair pair;
3934: ND red;
3935:
3936: s0 = 0; rp0 = 0; col = 0;
3937: while ( 1 ) {
3938: head = remove_head_pbucket_symbolic(bucket);
3939: if ( !head ) break;
3940: if ( !s0 ) s0 = head;
3941: else NEXT(s) = head;
3942: s = head;
3943: index = ndl_find_reducer(DL(head));
3944: if ( index >= 0 ) {
3945: h = nd_psh[index];
3946: NEWNM(mul);
3947: ndl_sub(DL(head),DL(h),DL(mul));
3948: if ( ndl_check_bound2(index,DL(mul)) ) return 0;
3949: MKNM_ind_pair(pair,mul,index);
3950: red = ndv_mul_nm_symbolic(mul,nd_ps[index]);
3951: add_pbucket_symbolic(bucket,nd_remove_head(red));
3952: NEXTNODE(rp0,rp); BDY(rp) = (pointer)pair;
3953: }
3954: col++;
3955: }
1.72 noro 3956: if ( rp0 ) NEXT(rp) = 0;
3957: NEXT(s) = 0;
1.65 noro 3958: s0v = (UINT *)MALLOC_ATOMIC(col*nd_wpd*sizeof(UINT));
3959: for ( i = 0, p = s0v, s = s0; i < col;
3960: i++, p += nd_wpd, s = NEXT(s) ) ndl_copy(DL(s),p);
3961: *s0vect = s0v;
3962: *r = rp0;
3963: return col;
3964: }
3965:
1.69 noro 3966: NODE nd_f4(int m)
3967: {
3968: int i,nh,stat,index;
3969: NODE r,g;
3970: ND_pairs d,l,t;
3971: ND spol,red;
3972: NDV nf,redv;
3973: NM s0,s;
1.74 ! noro 3974: NODE rp0,sp0,srp0,nflist;
1.69 noro 3975: int nsp,nred,col,rank,len,k,j,a;
3976: UINT c;
1.74 ! noro 3977: UINT **spmat;
1.69 noro 3978: UINT *s0vect,*svect,*p,*v;
3979: int *colstat;
3980: IndArray *imat;
3981: int *rhead;
3982: int spcol,sprow;
3983: int sugar;
3984: PGeoBucket bucket;
3985: struct oEGT eg0,eg1,eg_f4;
3986:
3987: if ( !m )
3988: error("nd_f4 : not implemented");
1.74 ! noro 3989: ndv_alloc = 0;
1.69 noro 3990: g = 0; d = 0;
3991: for ( i = 0; i < nd_psn; i++ ) {
3992: d = update_pairs(d,g,i);
3993: g = update_base(g,i);
3994: }
3995: while ( d ) {
3996: get_eg(&eg0);
3997: l = nd_minsugarp(d,&d);
3998: sugar = SG(l);
3999: bucket = create_pbucket();
1.74 ! noro 4000: stat = nd_sp_f4(m,l,bucket);
1.69 noro 4001: if ( !stat ) {
4002: for ( t = l; NEXT(t); t = NEXT(t) );
4003: NEXT(t) = d; d = l;
4004: d = nd_reconstruct(m,0,d);
4005: continue;
4006: }
4007: if ( !sp0 ) continue;
4008: col = nd_symbolic_preproc(bucket,&s0vect,&rp0);
4009: if ( !col ) {
4010: for ( t = l; NEXT(t); t = NEXT(t) );
4011: NEXT(t) = d; d = l;
4012: d = nd_reconstruct(m,0,d);
4013: continue;
4014: }
4015: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
1.74 ! noro 4016: if ( DP_Print )
! 4017: fprintf(asir_out,"sugar=%d,symb=%fsec,",
! 4018: sugar,eg_f4.exectime+eg_f4.gctime);
! 4019: if ( 1 )
! 4020: nflist = nd_f4_red(m,l,s0vect,col,rp0);
! 4021: else
! 4022: nflist = nd_f4_red_dist(m,l,s0vect,col,rp0);
1.69 noro 4023: /* adding new bases */
1.74 ! noro 4024: for ( r = nflist; r; r = NEXT(r) ) {
! 4025: nf = (NDV)BDY(r);
1.69 noro 4026: SG(nf) = sugar;
4027: ndv_removecont(m,nf);
4028: nh = ndv_newps(nf,0);
4029: d = update_pairs(d,g,nh);
4030: g = update_base(g,nh);
4031: }
4032: }
4033: for ( r = g; r; r = NEXT(r) ) BDY(r) = (pointer)nd_ps[(int)BDY(r)];
1.74 ! noro 4034: fprintf(asir_out,"ndv_alloc=%d\n",ndv_alloc);
1.69 noro 4035: return g;
4036: }
1.74 ! noro 4037:
! 4038: NODE nd_f4_red(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
1.63 noro 4039: {
1.67 noro 4040: IndArray *imat;
1.74 ! noro 4041: int nsp,nred,spcol,sprow,a;
1.65 noro 4042: int *rhead;
1.74 ! noro 4043: int i,j,k,l,rank;
! 4044: NODE rp,r0,r;
! 4045: ND_pairs sp;
! 4046: ND spol;
! 4047: int **spmat;
! 4048: UINT *svect,*v;
! 4049: int *colstat;
1.66 noro 4050: struct oEGT eg0,eg1,eg_f4;
1.74 ! noro 4051: NM_ind_pair *rvect;
1.63 noro 4052:
1.74 ! noro 4053: get_eg(&eg0);
! 4054: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
! 4055: nred = length(rp0); spcol = col-nred;
! 4056: imat = (IndArray *)ALLOCA(nred*sizeof(IndArray));
! 4057: rhead = (int *)ALLOCA(col*sizeof(int));
! 4058: for ( i = 0; i < col; i++ ) rhead[i] = 0;
1.63 noro 4059:
1.74 ! noro 4060: /* construction of index arrays */
! 4061: rvect = (NM_ind_pair *)ALLOCA(nred*sizeof(NM_ind_pair));
! 4062: for ( rp = rp0, i = 0; rp; i++, rp = NEXT(rp) ) {
! 4063: rvect[i] = (NM_ind_pair)BDY(rp);
! 4064: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rvect[i]);
! 4065: rhead[imat[i]->head] = 1;
! 4066: }
! 4067:
! 4068: /* elimination (1st step) */
! 4069: spmat = (int **)ALLOCA(nsp*sizeof(UINT *));
! 4070: svect = (UINT *)ALLOCA(col*sizeof(UINT));
! 4071: for ( a = sprow = 0, sp = sp0; a < nsp; a++, sp = NEXT(sp) ) {
! 4072: nd_sp(m,0,sp,&spol);
! 4073: nd_to_vect(m,s0vect,col,spol,svect);
! 4074: nd_free(spol);
! 4075: if ( m == -1 ) ndv_reduce_vect_sf(m,svect,col,imat,rvect,nred);
! 4076: else ndv_reduce_vect(m,svect,col,imat,rvect,nred);
! 4077: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
! 4078: if ( i < col ) {
! 4079: spmat[sprow] = v = (UINT *)MALLOC_ATOMIC(spcol*sizeof(UINT));
! 4080: for ( j = k = 0; j < col; j++ )
! 4081: if ( !rhead[j] ) v[k++] = svect[j];
! 4082: sprow++;
! 4083: }
! 4084: }
! 4085: /* free index arrays */
! 4086: for ( i = 0; i < nred; i++ ) GC_free(imat[i]->index.c);
! 4087:
! 4088: /* elimination (2nd step) */
! 4089: colstat = (int *)ALLOCA(spcol*sizeof(int));
! 4090: if ( m == -1 )
! 4091: rank = generic_gauss_elim_sf(spmat,sprow,spcol,m,colstat);
! 4092: else
! 4093: rank = generic_gauss_elim_mod(spmat,sprow,spcol,m,colstat);
! 4094: r0 = 0;
! 4095: for ( i = 0; i < rank; i++ ) {
! 4096: NEXTNODE(r0,r); BDY(r) =
! 4097: (pointer)vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
! 4098: GC_free(spmat[i]);
! 4099: }
! 4100: for ( ; i < sprow; i++ ) GC_free(spmat[i]);
! 4101: get_eg(&eg1); init_eg(&eg_f4); add_eg(&eg_f4,&eg0,&eg1);
! 4102: if ( DP_Print ) {
! 4103: fprintf(asir_out,"nsp=%d,nred=%d,spmat=(%d,%d),rank=%d ",
! 4104: nsp,nred,sprow,spcol,rank);
! 4105: fprintf(asir_out,"%fsec\n",eg_f4.exectime+eg_f4.gctime);
1.63 noro 4106: }
1.74 ! noro 4107: return r0;
! 4108: }
! 4109:
! 4110: FILE *nd_write,*nd_read;
! 4111:
! 4112: void nd_send_int(int a) {
! 4113: write_int(nd_write,&a);
! 4114: }
! 4115:
! 4116: void nd_send_intarray(int *p,int len) {
! 4117: write_intarray(nd_write,p,len);
! 4118: }
! 4119:
! 4120: int nd_recv_int() {
! 4121: int a;
! 4122:
! 4123: read_int(nd_read,&a);
! 4124: return a;
! 4125: }
! 4126:
! 4127: void nd_recv_intarray(int *p,int len) {
! 4128: read_intarray(nd_read,p,len);
! 4129: }
! 4130:
! 4131: void nd_send_ndv(NDV p) {
! 4132: int len,i;
! 4133: NMV m;
! 4134:
! 4135: if ( !p ) nd_send_int(0);
! 4136: else {
! 4137: len = LEN(p);
! 4138: nd_send_int(len);
! 4139: m = BDY(p);
! 4140: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
! 4141: nd_send_int(CM(m));
! 4142: nd_send_intarray(DL(m),nd_wpd);
1.68 noro 4143: }
1.74 ! noro 4144: }
! 4145: }
! 4146:
! 4147: void nd_send_nd(ND p) {
! 4148: int len,i;
! 4149: NM m;
! 4150:
! 4151: if ( !p ) nd_send_int(0);
! 4152: else {
! 4153: len = LEN(p);
! 4154: nd_send_int(len);
! 4155: m = BDY(p);
! 4156: for ( i = 0; i < len; i++, m = NEXT(m) ) {
! 4157: nd_send_int(CM(m));
! 4158: nd_send_intarray(DL(m),nd_wpd);
1.65 noro 4159: }
1.74 ! noro 4160: }
! 4161: }
1.65 noro 4162:
1.74 ! noro 4163: NDV nd_recv_ndv()
! 4164: {
! 4165: int len,i;
! 4166: NMV m,m0;
! 4167: NDV r;
1.65 noro 4168:
1.74 ! noro 4169: len = nd_recv_int();
! 4170: if ( !len ) return 0;
! 4171: else {
! 4172: m0 = m = (NMV)GC_malloc_atomic_ignore_off_page(nmv_adv*len);
! 4173: ndv_alloc += len*nmv_adv;
! 4174: for ( i = 0; i < len; i++, NMV_ADV(m) ) {
! 4175: CM(m) = nd_recv_int();
! 4176: nd_recv_intarray(DL(m),nd_wpd);
1.71 noro 4177: }
1.74 ! noro 4178: MKNDV(nd_nvar,m0,len,r);
! 4179: return r;
! 4180: }
! 4181: }
1.65 noro 4182:
1.74 ! noro 4183: int ox_exec_f4_red(Q proc)
! 4184: {
! 4185: Obj obj;
! 4186: STRING fname;
! 4187: NODE arg;
! 4188: int s;
! 4189: extern int ox_need_conv,ox_file_io;
! 4190:
! 4191: MKSTR(fname,"nd_exec_f4_red");
! 4192: arg = mknode(2,proc,fname);
! 4193: Pox_cmo_rpc(arg,&obj);
! 4194: s = get_ox_server_id(QTOS(proc));
! 4195: nd_write = iofp[s].out;
! 4196: nd_read = iofp[s].in;
! 4197: ox_need_conv = ox_file_io = 0;
! 4198: return s;
! 4199: }
! 4200:
! 4201: NODE nd_f4_red_dist(int m,ND_pairs sp0,UINT *s0vect,int col,NODE rp0)
! 4202: {
! 4203: int nsp,nred;
! 4204: int i,rank,s;
! 4205: NODE rp,r0,r;
! 4206: ND_pairs sp;
! 4207: NM_ind_pair pair;
! 4208: NMV nmv;
! 4209: NM nm;
! 4210: NDV nf;
! 4211: Obj proc,dmy;
! 4212:
! 4213: ox_launch_main(0,0,&proc);
! 4214: s = ox_exec_f4_red((Q)proc);
! 4215:
! 4216: nd_send_int(m);
! 4217: nd_send_int(nd_nvar);
! 4218: nd_send_int(nd_bpe);
! 4219: nd_send_int(nd_wpd);
! 4220: nd_send_int(nmv_adv);
! 4221:
! 4222: saveobj(nd_write,dp_current_spec.obj); fflush(nd_write);
! 4223:
! 4224: nd_send_int(nd_psn);
! 4225: for ( i = 0; i < nd_psn; i++ ) nd_send_ndv(nd_ps[i]);
! 4226:
! 4227: for ( sp = sp0, nsp = 0; sp; sp = NEXT(sp), nsp++ );
! 4228: nd_send_int(nsp);
! 4229: for ( i = 0, sp = sp0; i < nsp; i++, sp = NEXT(sp) ) {
! 4230: nd_send_int(sp->i1); nd_send_int(sp->i2);
! 4231: }
! 4232:
! 4233: nd_send_int(col); nd_send_intarray(s0vect,col*nd_wpd);
! 4234:
! 4235: nred = length(rp0); nd_send_int(nred);
! 4236: for ( i = 0, rp = rp0; i < nred; i++, rp = NEXT(rp) ) {
! 4237: pair = (NM_ind_pair)BDY(rp);
! 4238: nd_send_int(pair->index);
! 4239: nd_send_intarray(pair->mul->dl,nd_wpd);
! 4240: }
! 4241: fflush(nd_write);
! 4242: rank = nd_recv_int();
! 4243: fprintf(asir_out,"rank=%d\n",rank);
! 4244: r0 = 0;
! 4245: for ( i = 0; i < rank; i++ ) {
! 4246: nf = nd_recv_ndv();
! 4247: NEXTNODE(r0,r); BDY(r) = (pointer)nf;
! 4248: }
! 4249: Pox_shutdown(mknode(1,proc),&dmy);
! 4250: return r0;
! 4251: }
! 4252:
! 4253: /* server side */
! 4254:
! 4255: void nd_exec_f4_red_dist()
! 4256: {
! 4257: int m,i,nsp,col,s0size,nred,spcol,j,k;
! 4258: NM_ind_pair *rp0;
! 4259: NDV nf;
! 4260: UINT *s0vect;
! 4261: IndArray *imat;
! 4262: int *rhead;
! 4263: int **spmat;
! 4264: UINT *svect,*v;
! 4265: ND_pairs *sp0;
! 4266: int *colstat;
! 4267: int a,sprow,rank;
! 4268: struct order_spec ord;
! 4269: Obj ordspec;
! 4270: ND spol;
! 4271:
! 4272: nd_read = iofp[0].in;
! 4273: nd_write = iofp[0].out;
! 4274: m = nd_recv_int();
! 4275: nd_nvar = nd_recv_int();
! 4276: nd_bpe = nd_recv_int();
! 4277: nd_wpd = nd_recv_int();
! 4278: nmv_adv = nd_recv_int();
! 4279:
! 4280: loadobj(nd_read,&ordspec);
! 4281: create_order_spec(ordspec,&ord);
! 4282: nd_init_ord(&ord);
! 4283: nd_setup_parameters(nd_nvar,0);
! 4284:
! 4285: nd_psn = nd_recv_int();
! 4286: nd_ps = (NDV *)MALLOC(nd_psn*sizeof(NDV));
! 4287: nd_bound = (UINT **)MALLOC(nd_psn*sizeof(UINT *));
! 4288: for ( i = 0; i < nd_psn; i++ ) {
! 4289: nd_ps[i] = nd_recv_ndv();
! 4290: nd_bound[i] = ndv_compute_bound(nd_ps[i]);
! 4291: }
! 4292:
! 4293: nsp = nd_recv_int();
! 4294: sp0 = (ND_pairs *)MALLOC(nsp*sizeof(ND_pairs));
! 4295: for ( i = 0; i < nsp; i++ ) {
! 4296: NEWND_pairs(sp0[i]);
! 4297: sp0[i]->i1 = nd_recv_int(); sp0[i]->i2 = nd_recv_int();
! 4298: ndl_lcm(HDL(nd_ps[sp0[i]->i1]),HDL(nd_ps[sp0[i]->i2]),LCM(sp0[i]));
! 4299: }
! 4300:
! 4301: col = nd_recv_int();
! 4302: s0size = col*nd_wpd;
! 4303: s0vect = (UINT *)MALLOC(s0size*sizeof(UINT));
! 4304: nd_recv_intarray(s0vect,s0size);
! 4305:
! 4306: nred = nd_recv_int();
! 4307: rp0 = (NM_ind_pair *)MALLOC(nred*sizeof(NM_ind_pair));
! 4308: for ( i = 0; i < nred; i++ ) {
! 4309: rp0[i] = (NM_ind_pair)MALLOC(sizeof(struct oNM_ind_pair));
! 4310: rp0[i]->index = nd_recv_int();
! 4311: rp0[i]->mul = (NM)MALLOC(sizeof(struct oNM)+(nd_wpd-1)*sizeof(UINT));
! 4312: nd_recv_intarray(rp0[i]->mul->dl,nd_wpd);
! 4313: }
! 4314:
! 4315: spcol = col-nred;
! 4316: imat = (IndArray *)MALLOC(nred*sizeof(IndArray));
! 4317: rhead = (int *)MALLOC(col*sizeof(int));
! 4318: for ( i = 0; i < col; i++ ) rhead[i] = 0;
! 4319:
! 4320: /* construction of index arrays */
! 4321: for ( i = 0; i < nred; i++ ) {
! 4322: imat[i] = nm_ind_pair_to_vect_compress(m,s0vect,col,rp0[i]);
! 4323: rhead[imat[i]->head] = 1;
! 4324: }
! 4325:
! 4326: /* elimination (1st step) */
! 4327: spmat = (int **)MALLOC(nsp*sizeof(UINT *));
! 4328: svect = (UINT *)MALLOC(col*sizeof(UINT));
! 4329: for ( a = sprow = 0; a < nsp; a++ ) {
! 4330: nd_sp(m,0,sp0[a],&spol);
! 4331: nd_to_vect(m,s0vect,col,spol,svect);
! 4332: nd_free(spol);
! 4333: if ( m == -1 ) ndv_reduce_vect_sf(m,svect,col,imat,rp0,nred);
! 4334: else ndv_reduce_vect(m,svect,col,imat,rp0,nred);
! 4335: for ( i = 0; i < col; i++ ) if ( svect[i] ) break;
! 4336: if ( i < col ) {
! 4337: spmat[sprow] = v = (UINT *)MALLOC(spcol*sizeof(UINT));
! 4338: for ( j = k = 0; j < col; j++ )
! 4339: if ( !rhead[j] ) v[k++] = svect[j];
! 4340: sprow++;
! 4341: }
! 4342: }
! 4343: /* elimination (2nd step) */
! 4344: colstat = (int *)ALLOCA(spcol*sizeof(int));
! 4345: if ( m == -1 )
! 4346: rank = generic_gauss_elim_sf(spmat,sprow,spcol,m,colstat);
! 4347: else
! 4348: rank = generic_gauss_elim_mod(spmat,sprow,spcol,m,colstat);
! 4349: nd_send_int(rank);
! 4350: for ( i = 0; i < rank; i++ ) {
! 4351: nf = vect_to_ndv(spmat[i],spcol,col,rhead,s0vect);
! 4352: nd_send_ndv(nf);
1.63 noro 4353: }
1.74 ! noro 4354: fflush(nd_write);
1.59 noro 4355: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>